THE EFFICACY OF ANGLE-MATCHED ISOKINETIC KNEE FLEXOR AND EXTENSOR STRENGTH PARAMETERS IN PREDICTING AGILITY TEST PERFORMANCE

Abstract

Background

Agility is a fundamental performance element in many sports, but poses a high risk of injury. Hierarchical modelling has shown that eccentric hamstring strength is the primary determinant of agility performance.

Purpose

The purpose of this study was to investigate the relationship between knee flexor and extensor strength parameters and a battery of agility tests.

Study Design

Controlled laboratory study.

Methods

Nineteen recreational intermittent games players completed an agility battery and isokinetic testing of the eccentric knee flexors (eccH) and concentric knee extensors (conQ) at 60, 180 and 300°·s⁻¹. Peak torque and the angle at which peak torque occurred were calculated for eccH and conQ at each speed. Dynamic control ratios (eccH:conQ) and fast:slow ratios (300:60) were calculated using peak torque values, and again using angle-matched data, for eccH and conQ. The agility test battery differentiated linear vs directional changes and prescriptive vs reactive tasks.

Results

Linear regression showed that eccH parameters were generally a better predictor of agility performance than conQ parameters. Stepwise regression showed that only angle-matched strength ratios contributed to the prediction of each agility test. Trdaitionally calculated strength ratios using peak torque values failed to predict performance. Angle-matched strength parameters were able to account for 80% of the variation in T-test performance, 70% of deceleration distance, 55% of 10m sprint performance, and 44% of reactive change of direction speed.

Conclusions

Traditionally calculated strength ratios failed to predict agility performance, whereas angle-matched strength ratios had better predictive ability and featured in a predictive stepwise model for each agility task.

Level of Evidence

2c

INTRODUCTION

The ability to change direction rapidly has implications for both performance¹ and injury risk.² A hierarchical ordering of anthropometric, cognitive, and knee flexor/extensor strength parameters across a battery of agility tests showed that eccentric hamstring strength was the primary determinant of success.³ To date, the majority of research investigating the predictors of agility performance has focused on markers of leg strength.^4,5 This focus within the literature reflects the functional role of eccentric hamstring strength in enhancing neuromuscular control during the foot ground contact phase of cutting activities.⁶

Strength parameters, as derived from isokinetic testing, have traditionally included peak torque, angle of peak torque, and ipsi-lateral strength ratios. The use of peak concentric quadriceps and concentric hamstring strength was usurped by the ‘functional ratio’,⁷ which quantifies the ratio of peak eccentric hamstring (eccH) torque to peak concentric quadriceps (conQ) torque. This ratio has greater mechanical specificity to reflect the reciprocal antagonistic muscular function during core skills such as running. However, the use of peak values negates that the quadriceps and hamstrings will exhibit their peak torque at different joint angles,⁸ and recently angle-matched data have been advocated as a means of providing a more functionally relevant measure.^9,10

The purpose of the present study was to investigate the relationship between knee flexor and extensor strength parameters and a battery of agility tests. The test battery was designed to incorporate the varied elements of agility, and therefore to differentiate between linear speed, prescriptive and reactive change of direction, and deceleration.³ The range of components influencing agility performance lends itself to a hierarchical ordering of factors, and in this study both the traditional and angle-matched derivations of strength ratios are considered. It was hypothesized that the angle-matched strength ratios would provide a stronger predictor of athletic performance.

METHODS

Participants

Nineteen male intermittent team sports players (mean ± S.D.; age: 22.1 ± 1.9 years; height: 182.9 ± 5.5 cm; body mass: 77 ± 4.9 kg) who competed in rugby or soccer completed the study. Intermittent team sport players were recruited so that all participants were familiar with the functional challenges posed by the test battery. Additional inclusion criteria required players to be injury free for three months preceding data collection. Participants provided written informed consent with ethical approval granted at the host university departmental ethics committees, and in the spirit of the Helsinki Declaration. No potential conflicts of interest were noted. All testing was conducted between 14:00–16:00 hours in accord with regular competitive demands of these players, and to negate the influence of circadian effects on performance.

Agility testing battery

The testing battery was comprised of the following: a T-test, a 10 m linear sprint, a reactive change of direction cutting task, and a reactive deceleration task, performed in randomized order and as described in previous applications.³ This battery (Figure 1) was designed to differentiate between prescriptive vs. reactive tasks, linear vs. multi-directional speed, and acceleration vs. deceleration. All agility tests were completed using commercially available photoelectric timing gates (Smartspeed, Fusion Sport, Australia) with testing preceded by a dynamic warm-up and a demonstration of each agility test.

Figure 1.

The agility testing battery.

Isokinetic profiling

All subjects completed isokinetic dynamometry assessments within ±2 weeks of the agility testing (System 3, Biodex Medical Systems, Shirley, NY, USA). Testing comprised of eccentric knee flexor (in reactive eccentric mode) and concentric knee extensor strength at angular velocities of 180, 300, and 60°·s⁻¹.¹¹ The dynamometer setup was modified so as to be subject specific, following the manufacturer's guidelines. The crank axis was aligned with the axis of rotation of the knee joint, and the cuff of the dynamometer's lever arm was secured around the ankle, proximal to the malleoli. With the subject in the seated position, restraints were applied across the test thigh, proximal to the knee joint so as not to restrict movement, and across the chest. Familiarization trials were completed in each mode and at each speed, with data collection comprising five maximal contractions at each speed.¹² The recovery phase between maximal efforts was set as passive knee flexion at 10°·s⁻¹, requiring no exertion from the subject. There was an allocated 90 seconds rest between sets. Communication to each participant was restricted to informing them of the test speeds with no visual feedback offered.

Peak torque (T) and angle of peak torque (θ) were calculated at each speed for both eccH and conQ. The dynamic control ratio (DCR), defined as eccH:conQ was calculated at each test speed using peak torque values. Fast:slow (300:60) ratios were also calculated for both eccH and conQ using peak torque values. The ipsi-lateral eccH:conQ ratio, and the fast:slow ratio were also calculated using angle-matched torque data. All calculations were restricted to the isokinetic phase of movement.

Statistical Analyses

Agility test performance data and all isokinetic parameters are quantified as mean ± standard deviation. Linear regression analysis was used to quantify the relationship between performance on each agility test with peak torque, and with angle of peak torque at each discrete testing speed (60, 180, 300 º·s⁻¹), and for each modality (eccH, conQ). Multiple linear regression analysis was then used to model agility test performance as a function of peak torque across all testing speeds collectively, for eccH and for conQ. This process was repeated for angle of peak torque across all speeds in each modality. In all cases the correlation coefficient (r) was used to quantify the relative contribution of each factor to agility performance. The value r² was subsequently calculated to quantify the percentage variation in agility performance that can be accounted for by variation in the isokinetic variable.

Finally, and in order to develop a hierarchical ordering of the strength parameters influencing each agility test, a forward stepwise regression model was utilized. Stepwise linear regression provides a means of including multiple variables within a model while simultaneously removing those variables that are not important. The forward selection model employed is initiated with no variables included, and subsequently adding the variable whose insertion gives the most statistically significant improvement of the correlation. This process is repeated until no additional variables improve the model to a statistically significant extent. This process allowed for identification of the singular most important isokinetic contributor to agility test performance, defined by the greatest magnitude of r. Additional strength parameters were added in sequence, only if their insertion improved the magnitude of r. The number of model steps is therefore unique to each agility test, and each step is quantified by the correlation coefficient at that level.

RESULTS

Table 1 summarizes the performance for each agility test, and for each isokinetic analysis parameter. Peak torque and angle of peak torque are expressed for eccH and conQ at each testing speed. The dynamic control ratio and fast:slow strength ratios are provided for both traditional and angle-specific methods. Angle-matched strength ratios are distinguished with the use of a superscript θ.

Table 1.

Agility test and isokinetic strength performance of 19 male intermittent team sport players.

10m Sprint t10 (sec)	T-test tT (sec)	Reactive Cut tRC (sec)	Deceleration d (m)
1.77 ± 0.15	10.77 ± 0.73	2.69 ± 0.21	5.59 ± 0.52
Peak Torque (Nm)	eccH T60	eccH T180	eccH T300
	139.61 ± 33.82	165.50 ± 30.24	164.86 ± 26.60
	conQ T60	conQ T180	conQ T300
	203.24 ± 35.64	148.55 ± 32.07	121.99 ± 25.54
Angle of Peak Torque (º)	eccH θ60	eccH θ180	eccH θ300
	51.53 ± 12.44	47.89 ± 12.86	59.79 ± 13.21
	conQ θ60	conQ θ180	conQ θ300
	73.84 ± 4.78	67.74 ± 5.58	60.58 ± 5.29
Dynamic Control Ratio	DCR60	DCR180	DCR300
	0.70 ± 0.16	1.18 ± 0.42	1.39 ± 0.29
	${\text{DCR}}_{60}^{\theta }$	${\text{DCR}}_{180}^{\theta }$	${\text{DCR}}_{300}^{\theta }$
	0.86 ± 0.29	1.27 ± 0.33	1.44 ± 0.31
Fast:Slow Ratio	EccH fast:slow	EccHθ fast:slow
	1.23 ± 0.20	1.20 ± 0.25
	ConQ fast:slow	ConQθ fast:slow
	0.67 ± 0.25	0.74 ± 0.21

Table 2 quantifies r² for the squarelinear correlation of agility test performance as a function (f) of each parameter determined for eccH and conQ. Individual correlation coefficients in eccH peak torque (T) were lowest in the Reactive Cut (r² ≤ 0.02), and typically greatest in the T-test at all testing speeds. The single highest correlation strength was observed for eccH T₁₈₀ (r² = 0.61). Individual correlation coefficients in T were lower for conQ than eccH in all tests with the exception of the Reactive Cut, although values were still small (r² ≤ 0.10). Correlation coefficients in the angle of peak torque (θ) were also highest for eccH in the T-test (r² ≤ 0.27), except at the fastest testing speed. As with peak torque, the strength of the individual correlation coefficients was typically greater in eccH than for conQ. When peak torque was considered across all testing speeds, eccH strength was able to account for up to 62% of variation in T-test performance, but only 2% of Reactive Cut performance. The angle of peak torque across all speeds was able to account for between 7% (10m Sprint) and 44% (T-test) of agility performance in eccH. These values were lower for conQ, with collective peak torque accounting for between 9% (Deceleration) and 24% (T-test) of variation in performance. Angle of peak conQ torque across all speeds was also greatest for T-test (r² = 0.16), with only 2% of Deceleration performance accounted for.

Table 2.

Single and multiple linear regression analyses to quantify the correlation (r²) between isokinetic strength parameters and agility test performance.

		10m Sprint t10 (sec)	T-test tT (sec)	Reactive Cut tRC (sec)	Deceleration d (m)
eccH	ƒ (T60)	0.11	0.32	0.01	0.32
	ƒ (T180)	0.23	0.61	0.01	0.21
	ƒ (T300)	0.21	0.41	0.02	0.22
	ƒ (T60,T180,T300)	0.26	0.62	0.04	0.33
	ƒ θ60	0.05	0.26	0.01	0.11
	ƒ θ180	0.03	0.27	0.07	0.12
	ƒ θ300	0.01	0.01	0.01	0.02
	ƒ (θ 60, θ 180, θ 300)	0.07	0.44	0.10	0.20
	ƒ (T60,T180,T300, θ 60, θ 180, θ 300)	0.33	0.75	0.16	0.45
conQ	ƒ (T60)	0.03	0.04	0.07	0.01
	ƒ (T180)	0.01	0.01	0.10	0.01
	ƒ (T300)	0.01	0.07	0.10	0.02
	ƒ (T60,T180,T300)	0.10	0.24	0.12	0.09
	ƒ θ60	0.01	0.01	0.05	0.01
	ƒ θ180	0.04	0.14	0.03	0.01
	ƒ θ300	0.10	0.01	0.01	0.01
	ƒ (θ 60, θ 180, θ 300)	0.14	0.16	0.09	0.02
	ƒ (T60,T180,T300, θ 60, θ 180, θ 300)	0.18	0.36	0.20	0.09

eccH = eccentric knee flexor; conQ = concentric knee extensor;

T = peak torque; θ = angle of peak torque;

60, 180, 300 = isokinetic testing speeds in º·s⁻¹; ƒ = as a function of

Table 2 also summarizes a multiple linear regression to quantify the predictive potential of all parameters in peak torque (and at all speeds). With all variables considered, eccH proved a stronger determinant of performance than conQ in the 10m sprint (eccH r² = 0.33; conQ r² = 0.18), T-test (eccH r² = 0.75; conQ r² = 0.36), and Deceleration test (eccH r² = 0.45; conQ r² = 0.09). While the weakest correlations were observed in the Reactive Cut (eccH r² = 0.16; conQ r² = 0.20), the greatest correlation was observed between eccentric hamstring strength parameters and T-test performance (r² = 0.75, p = 0.01).

The previous data includes only parameters relating to peak torque, with no consideration of angle-matched data. In the next series of regressions the strength parameters were considered as both traditional (using peak torque values), and angle-matched. Table 3 presents the hierarchical ordering of factors influencing each agility tests. The original data set included peak torque, angle of peak torque, eccH:conQ and fast:slow ratios (using peak torque and angle-matched data). Angle-matched strength ratios are distinguished with the use of a superscript θ. Stepwise modelling produced a hierarchical model of determinants, with isokinetic parameters able to account for between 44% (Reactive Cut) and 80% (T-test) of the variation in agility test performance. In the 10m Sprint test the hierarchical model was comprised entirely of variables from the eccH modality. T₁₈₀ was the primary predictor of 10m Sprint performance, with the fast:slow ratio and θ₃₀₀ subsequently added. The T-test also had eccH T₁₈₀ as the primary predictor, with the high speed DCR as a secondary element. Both additional elements added to this model were also in eccH, with θ₆₀ and the fast:slow ratio producing a cumulative r² = 0.80. The high speed DCR was the primary predictor of Reactive Cut performance, where slow conQ θ₆₀ also appeared in the model as a tertirary predictor. EccH elements also dominated Deceleration performance, with T₆₀, the fast:slow ratio and θ₆₀ the first elements to be included, and θ₁₈₀ in conQ the final addition to produce a total r² of 0.70.

Table 3.

A hierarchical linear regression model of isokinetic strength factors influencing agility test performance using a forward stepwise approach.

	10m Sprint t10 (sec)	T-test tT (sec)	Reactive Cut tRC (sec)	Deceleration d (m)
Step 1	eccH T180 r2 = 0.23	eccH T180 r2 = 0.61	DCR300 r2 = 0.19	eccH T60 r2 = 0.32
Step 2	eccH fast:slow r2 = 0.42	DCR300 r2 = 0.70	eccH θ180 r2 = 0.32	eccH fast:slow r2 = 0.53
Step 3	eccH θ300 r2 = 0.55	eccH θ60 r2 = 0.76	conQ θ60 r2 = 0.44	eccH θ60 r2 = 0.62
Step 4		eccH fast:slow r2 = 0.80		conQ θ180 r2 = 0.70

eccH = eccentric knee flexor; conQ = concentric knee extensor;

T = peak torque; θ = angle of peak torque; DCR = Dynamic control ratio using peak torque values; 60, 180, 300 = isokinetic testing speeds in º·s⁻¹

DISCUSSION

The aim of the present study was to investigate the relationship between knee flexor and extensor strength measures and an agility testing battery. The agility tests were designed to differentiate between linear and change of direction speed, prescriptive and reactive drills, acceleration and deceleration. Additional rigour was provided in measures of isokinetic strength with recent research advocating the use of angle-matched analyses to more closely replicate the functional kinesiology of such tasks.^9,10

Considering each agility test as a function of peak torque and angle at peak torque at all test speeds, eccentric hamstring strength proved a stronger predictor of performance in three of the tests and was shown to account for 75% of the variance in the T-test. The hamstrings musculature enhances functionality in both linear and directional change tasks,¹³ helping to maintain hip extensor torque, assisting dynamic trunk stabilisation, and controlling knee flexion.⁶ The functional role of the hamstrings in neuromuscular control during the ground contact phase enhances the development of stride frequency,⁶ and thus the multiple changes of speed and/or direction in the T-test makes most functional use of the hamstrings. In contrast, the lack of relationship with the reactive cut might be attributed to insufficient time to generate muscle force.¹⁴ The greater relative contribution to predictive change of direction speed is in line with previous work, where eccentric hamstring strength was able to discriminate between the best and worst T-test performers.¹⁵

The strong correlation between T-test performance and eccentric hamstring strength was further analyzed using a forward stepwise regression to develop a hierarchical model of each test. Eccentric hamstring strength at 180°·s⁻¹ was the standout contributor, accounting for 61% of the variability in task performance. Peak torque at this mid-range speed was also the primary predictor of 10m sprint (r² = 0.23, P = 0.24) performance. In contrast, peak torque at the slowest speed of 60°·s⁻¹ best predicted deceleration task performance (r² = 0.32, P = 0.01). This relatively slow isokinetic test speed is most often used in literature considering the correlation between agility and isokinetic strength,^6,15 but the present study highlights the task-specific functional demands of strength and the necessity to conduct isokinetic profiles across a range of speeds. In summation, based on this agility battery, slow speed strength best predicts deceleration performance while moderate speed strength is more srongly correlated with tasks emphasising acceleration. This functional relationship has implications for both training and (p)rehabilitation.

The hierarchical modelling of factors affecting agility performance failed to include strength ratios calculated using the traditional peak torque values. The limitation of this dynamic control ratio is the assumption that peak eccentric hamstring and concentric quadriceps torque are co-dependent.⁸ In contrast, the angle-matched strength ratios were a more powerful predictor of agility performance. The primary contribution from eccentric hamstring strength in predicting agility test performance was supplemented by angle-specific strength ratios. The high speed angle matched dynamic control ratio was the first or second placed parameter in a hierarchical ordering of both the reactive and prescriptive change of direction tasks respectively. In contrast, the linear tasks both exhibited the angle matched fast:slow eccentric hamstring strength ratio as a second predictor of performance. The angle matched fast:slow ratio for eccentric hamstring strength was included in the full hierarchical modelling of three of the agility tests. The impact of the angle-matched dynamic control ratio in change of direction tasks highlights the greater functional relevance, and subsequently greater correlation with athletic performance than traditional strength ratios based on peak torque and independent of joint angle.

This stepwise process created a multiple linear regression equation for each agility test, with up to 80% of the variability in performance accounted for. The strongest correlation for the T-test probably reflects the frequency and variety of functional demands of this test, which includes the discrete elements of the other tests in linear acceleration, deceleration, and change of direction. The weakest correlation was observed for the reactive cutting task, most likely reflecting the greater contribution to this test from the psychological strand of the agility model,^3,16 which was not considered in the present study. Comparisons with previous literature are limited, slow eccentric knee flexor strength explaining 39% of variance in a change of direction task⁶ and in a multi-factorial investigation including slow eccentric hamstring strength accounting for ∼45% of the variance in the predictive T-test and 5m shuttle run test.¹⁵ The higher values reported in the current study are encouraging and reflect a more functional approach to isokinetic profiling, particularly in relation to the testing speed which incorporated profiling up the maximum available 300°·s⁻¹. The inclusion of higher speed analyses also enables the use of fast:slow strength ratios, which (when using angle matched data) were consistent factors influencing agility task performance. Given the functional role of the hamstring musculature during such tasks,¹⁷ the force-velocity relationship warrants greater consideration in conditioning, training the hamstring musculature closer to speeds with functional relevance to the task.

The multiple regression coefficients presented in Table 3 are such that the variability in task performance not explained by the strength parameters not included in the present study ranged from 20% in the T-test to 56% in the reactive cut. This highlights the complexity in determining those factors affecting agility, and particularly the ‘perceptual and decision-making factors’ fundamental in reactive agility.¹⁶ Incorporating both physical and psychological strands of the agility model into a hierarchical ordering of varying agility tasks would therefore have merit. There is also scope for greater consideration of technique, and it must be acknowledged that the nature and size of the population used in the present study also limits generalization. In the present study soccer and rugby players were grouped as being representative of intermittent team sports. Future research might consider a more rigorous grouping of participants, given potential differences in the agility demands of these sports. Gender disparities in the frequency and severity of anterior cruciate ligament injury ² and the relevance of agility tasks to common injury mechanisms would also suggest that a comparison of male and female athletes would be beneficial. Exclusion crieteria employed in the current study negated the opportunity to further examine this relationship in participants with a history of knee or hamstring injury. The analysis extended to participants representing differences in gender, level of competition, sport, and injury history would all be valuable contributions. The epidemiological observations of increased injury risk during the latter stages of sports also warrant investigation of whether these correlations might change with fatigue. It must therefore be acknowledged that interpretation of the data presented in the current study should not be generalized beyond a male, non-elite intermittent team sports player, with no previous injury history.

CONCLUSIONS

Weak correlations across the agility testing battery highlight the varied and distinct technical elements of each test, such that in profiling agility a battery of tests is advocated. Eccentric hamstring strength typically displayed a stronger correlation coefficeient with agility performance than concentric quadriceps strength, highlighting the functional role of the hamstrings in such tasks. Strength ratios based on peak torque values failed to predict agility performance, whereas angle-matched strength ratios featured prominenetly in predictive stepwise models of performance. The speed of eccentric hamstring strength was also an important factor, advocating profiling at faster, more functional speeds. A combination of strength parameters was able to predict up to 78% of agility performance, but the efficacy and strength of the prediction is task dependent. The use of ipsi-lateral strength ratios has merit, but the fast:slow ratio warrants inclusion also. With a comprehensive isokinetic profile, functionally matched to the athletic task, there is efficacy in predicting agility test performance using isokinetic profiling.