Countermovement Jump and Isokinetic Dynamometry as Measures of Rehabilitation Status After Anterior Cruciate Ligament Reconstruction

This study was carried out in an orthopaedic private hospital at which more than 700 ACL surgical reconstructions take place each year. The hospital's research and ethics committee approved this study (25-AFM-010), which was registered at Clinicaltrials.gov (NCT02771548).

All male multidirectional field-sport athletes aged 24 to 26 years who attended a surgical consultation at least 6 months after ACLR were invited to participate. The ACLR group comprised 118 patients, from October 2013 to March 2014, who met the inclusion criteria and gave informed consent. Inclusion criteria were being male and having undergone primary ACLR surgery using a patellar-tendon graft carried out by a single surgeon. Excluded were individuals who met any of the following criteria: being female or having experienced a concomitant knee fracture or undergone other knee-ligament reconstruction, revision of a previous ACLR, or a previous contralateral ACLR.

Healthy male participants (n = 44) from a multidirectional sport with no history of lower limb injury were recruited as controls. For the purpose of this study, multidirectional sports included organized sports that require movements in multiple planes and multiple directions that can be random and reactive in nature—forward, backward, and side to side.

An 8-camera motion-analysis system (200 Hz; model Bonita B10; Vicon Motion Systems, Hauppauge, NY), synchronized with two 40- × 60-cm force platforms (1000 Hz; model BP400600; Advanced Mechanical Technology Inc, Watertown, MA), was used to collect the kinematic and kinetic data of the SL CMJ. Data were sampled to 200 Hz, and the Vicon Plug-in-Gait marker set was used as described by Marshall et al.³¹  The inverse-dynamics approach was used to calculate joint kinetics.³¹ 

The athlete's height (m), mass (kg), and operated limb were recorded before testing began. The SL CMJ was assessed first, followed by ISO, and the nonoperated limb was tested first for both. Before jump testing, participants performed a standardized warm-up that consisted of 5 double-legged squats and 2 submaximal double-legged CMJs, followed by 2 submaximal SL CMJs on the leg to be tested. The participants were instructed to stand with 1 foot on the force plate and the free leg behind at approximately 90°. With their hands on their iliac crests, they were asked to complete an SL CMJ, jumping as high as possible. The SL CMJ has high test-retest reliability in patients with ACLR (intraclass correlation coefficient = 0.86–0.97).³² 

Although 3 trials were recorded on each leg, we analyzed only the highest jump. Jump height was determined using the impulse-momentum relationship.¹¹  Jump height represents an index of muscle power that is independent of body size; thus, the recorded heights (cm) of vertical jumps were not normalized to body mass.³³  Maximum power (peak power) was measured and normalized to body weight during the propulsion phase of the SL CMJ. Power generated at the hip, knee, and ankle joint in the sagittal plane was expressed as a percentage of total peak power (sum of all 3 joints) during the propulsion phase of the SL CMJ. We chose peak power because, in contrast to peak moment, peak power has a strong relationship (r > 0.72) with jump height.^34,35 

After jump testing was conducted, concentric knee-extension and -flexion peak torque were assessed at an angular velocity of 60°/s using an ISO dynamometer (model Cybex Norm; Computer Sports Medicine Inc, Stoughton, MA). High relative reliability and moderate absolute reliability have been found for measuring peak-torque values of concentric knee extension at 60°/s using the Cybex Norm.³⁶  The participants first performed a warm-up set that consisted of 5 repetitions of extension and flexion at 50% to 75% and 1 attempt at 100% of maximal effort. After a 60-second rest period, the athletes were then instructed to complete 2 maximal-extension and -flexion sets of 5 repetitions. They were instructed to push and pull as hard and fast as possible against the resistance. The procedure was then repeated on the operated limb. The set with the lowest coefficient of variance was analyzed. Knee-extension and -flexion peak torques normalized to body mass (Nm·kg^–1) and LSI were the primary variables of interest.

Descriptive statistics are presented as mean ± standard deviation and 95% confidence interval (CI) where applicable. Paired-samples t tests were carried out to detect statistical differences between the operated and nonoperated limbs, and an independent-samples t test was conducted between groups. When testing between groups for differences in non-LSI measures, the operated limb of the ACLR group was compared with the dominant limb (kicking leg) of the control group. The SL CMJ height (cm), peak power, joint power contributions, LSI, and LSI_modified were compared between limbs and groups. The ISO variables analyzed were knee-extension and -flexion peak torque as a percentage of body mass, LSI, and LSI_modified; LSI and LSI_modified were calculated according to equations 1 and 2 shown earlier. To examine the effect of statistical differences, the Cohen d was calculated and classified as small (0.20–0.49), moderate (0.50–0.79), or large (≥0.80).³⁷  To examine the probability of committing a type II error during the analysis and find the smallest statistically significant difference that would have been detectable with the given sample size, we performed a post hoc power analysis.

A stepwise logistic regression enabled us to identify the combination of variables, using features of (1) the ISO test, (2) the SL CMJ, and (3) both exercises to most accurately model the response variable (ACLR = 0 or control = 1). To robustly examine which features should be included in a regression model, we repeated the stepwise logistic regression process 100 times using different randomly selected participants' training data (108 ACLR and 34 control). Features that were chosen more than 50 times were used as input variables for a multinomial regression model, which was fit to a randomized training set (108 ACLR and 34 control) and subsequently used to predict the group membership of a testing set (10 ACLR and 10 control). The accuracy of the model was assessed by comparing predicted to actual membership. The process was repeated 100 times using different randomly selected participants' training data and testing sets to achieve a robust measure of the expected accuracy. Coefficients and intercepts were averaged to describe the prediction function.

Significance was set at P < .05. Data processing and descriptive statistics were carried out using MATLAB (version R2015a; The MathWorks Inc, Natick, MA).

Demographics did not differ between the healthy and ACLR cohorts (Table 1). The ACLR participants had undergone surgery an average of 6.6 ± 1.0 months earlier. The statistical power was 0.95 or greater for all significantly different measures, with effect sizes greater than 0.5.

Between limbs of the ACLR group, all ISO and SL CMJ measures were different except for the percentage of ankle power contribution. With the exception of the percentage of hip power contribution, which was higher, the operated limb demonstrated smaller values (Table 2). Based on effect size, the order of the differences from highest to lowest was ISO knee-extension peak torque (d = –1.33), SL CMJ hip power contribution (d = 0.75), SL CMJ height (d = –0.71), SL CMJ peak power (d = –0.47, β = 0.99), SL CMJ knee power contribution (d = –0.37, β = 0.97), and ISO knee-flexion peak torque (d = –0.19, β = 0.54).

In the control group, only 2 measures were different. The SL CMJ peak power was higher for the nondominant limb (d = –0.26, β = 0.41), while the ISO knee-extension peak torque was larger for the dominant limb (d = 0.20, β = 0.25). The control group displayed no differences between limbs at the hip, knee, or ankle (P > .05, Cohen d < 0.15) in proportion of power generation, with a similar distribution evident in the nonoperative limb of the ACLR group.

Between-groups differences were significant for all ISO measures except for knee-flexion peak torque LSI_modified. Based on effect size, the order of the differences from highest to lowest was ISO knee-extension LSI (d = –1.53), LSI_modified (d = 1.28), ISO knee-extension peak torque (d = –1.20), SL CMJ height LSI (d = –1.12), SL CMJ height (d = –0.86), SL CMJ peak power LSI_modified (d = –0.61), hip power contribution (d = 0.61), SL CMJ knee power contribution (d = –0.40, β = 0.70), and ISO knee-flexion peak torque (d = –0.36, β = 0.64).

The stepwise logistic regression identified the following variables as most appropriate for predicting group membership: percentage of hip power contribution, ISO knee-extension peak torque, knee-extension LSI_modified, and SL CMJ height LSI_modified. A detailed description can be found in Table 3. The results of the logistic regression suggest that a combination of variables from ISO and SL CMJ testing can best predict group membership (89.2%), followed by the ISO model (84.6%) and the SL CMJ value (74.7%).

Normative values and 95% CIs are reported for the ISO and SL CMJ variables in Table 4. In the control group, ISO knee-extension peak torque was 261% of body mass in the dominant and 253% in the nondominant limb. The ISO knee-flexion peak torque was a mean of 155% of body mass in both limbs. Healthy participants jumped a mean of 17 cm using either limb. The distribution of extension power generation in the control group (rounding to the nearest 10%) was 20% hip, 30% knee, and 50% ankle bilaterally. Limb symmetry indices were approximately 100% for SL CMJ height, 96% for peak power, and 104% for ISO knee-extension and 101% for ISO knee-flexion peak torques, with standard deviations of approximately 10% for these healthy participants.

The normative (control) cohort differed strongly from the ACLR group in ISO knee-extension LSI, ISO knee-extension LSI_modified, ISO knee-extension peak torque, SL CMJ height LSI, and SL CMJ height, whereas the combination of proportion of power at the hip, ISO knee-extension peak torque, knee-extension LSI_modified, and SL CMJ height LSI_modified identified group membership with an accuracy of 89%. These results may help us to determine which tests and variables most accurately identify rehabilitation deficits after ACLR.

The ACLR group demonstrated large limb asymmetries in ISO knee-extension peak torque, while the control group displayed small but significant asymmetries. The effect size between limbs in the control group (d = 0.20) was small.³⁷  The largest effect size was between groups in ISO knee-extension peak torque for magnitude, LSI, and LSI_modified. Small but significant differences were present between groups in ISO knee-flexion peak torque variables, which will not be discussed due to its high probability of type II error (β = 0.64).

The ISO knee-extension peak-torque LSI demonstrated the largest effect size and, accordingly, was the most sensitive measure to differences between the ACLR and control groups. Interestingly, the control group also demonstrated a difference between limbs in knee-extension peak torque and symmetry, so it may be important to look at other variables when assessing readiness to return to play. These findings confirm the value of examining knee-extension peak-torque symmetry as a measure of rehabilitation status and may be used to guide ACLR rehabilitation.^9,15,17 

The ACLR group demonstrated large limb asymmetries in SL CMJ height, whereas no difference was found between limbs in the control group. Large effect sizes were present between groups for SL CMJ height, LSI, and LSI_modified. Moderate differences occurred between groups for SL CMJ peak power. Control participants displayed symmetry in jump height and joint power generated in both limbs, indicating that symmetric measures will be important in assessing completion of rehabilitation.

The athletes with ACLR differed from healthy athletes in the percentage of power per joint. A higher proportion of peak power was generated through the hip in the ACLR group's operated limb compared with both their nonoperated limbs and the control group. This reduction at the knee suggests a redistribution of effort to the hip during the SL CMJ in the operated limbs. Joint power generation at the knee was previously shown¹⁹  to be reduced in the stance phase of running, which suggests that after ACLR, load distribution may be altered during functional movements.

The difference in power distribution may be related to the lower strength values identified in the ACLR group compared with their healthy counterparts. We conducted a post hoc Pearson correlation to examine the relationship between ISO strength and the power generation at each joint in the ACLR and control groups (Table 5). The distribution of power generation across the hip, knee, and ankle joints was positively correlated with ISO knee-extension peak torque in the ACLR group (Pearson r = 0.28–0.31, P < .002), but no significant correlation was found between these factors in the control group (P > .05). This may indicate that strength has a small association with extension power generation distribution across the lower limb joints after ACLR.

Our findings suggest that operated limbs may compensate for lower peak power generation at the knee by generating a higher proportion of power at the hip during jumping. Previous investigators³⁸  determined that the hip or ankle extensors compensated for a knee-extension deficit in vertical jump landing, where differences between operated and nonoperated limbs may predispose people to a risk of injury. It is important to note, however, that participants in their study completed a CMJ with arm swing, whereas those in our study did not, potentially influencing the jumping and landing mechanics.

The stepwise logistic regression identified the ISO measures of knee-extension peak torque relative to body mass and LSI_modified as well as jump height and proportion of power generated at the hip in SL CMJ as the most important predictors of ACLR- or control-group membership (89% accuracy; Table 3). With ISO measures alone, the accuracy was 84.6%; with SL CMJ measures alone, the accuracy was 74.7%.

Although ISO knee-extension LSI had the largest between-groups difference (Cohen d = –1.53), the regression analysis selected its modified version. One possible reason might be that the LSI measurement does not account for operated, nonoperated, dominant, or nondominant limb, which could have artificially inflated the effect sizes for the LSI values. Such effects may have been removed due to the repetitive regression analysis using different datasets.

In this model, the first selected predictor was percentage of hip power generation in jumping. Assessing single-plane ISO strength in a single joint and a single muscle group, such as the quadriceps, is important for identifying a fully rehabilitated athlete; however, assessing functional tasks allows investigators to observe movement in a dynamic manner, similar to the activities anticipated on return to play.

We acknowledge that although strength and power are important, these 2 factors alone do not tell the whole story. This logistic regression did not take into account motor control at multiple joints of the lower limb or the interaction of the trunk during the jumping task. Screening tools such as the Landing Error Scoring System³⁹  and tuck jump¹⁵  are inexpensive and easy-to-use field tests that aim to identify this compensation at the hip by looking for poor movement patterns. However, quantifying the limitation in joint power and strength may enhance the assessment of rehabilitation status after ACLR and add to the specific, measurable, and individualized targets and goals that patients can aim to achieve postoperatively.

Absolute values were recorded for the control group in this study, allowing estimation of appropriate targets for male patients in the 24- to 26-year-old age group after ACLR. Based on the reported 95% CIs, these participants should be able to jump between 16 cm and 18 cm to be considered recovered. They should achieve 240% to 270% of their body mass on ISO knee-extension testing and 150% to 160% of their body mass on ISO knee-flexion testing. These results also confirm accepted practice with regard to symmetry: aiming for greater than 90% of the opposite limb in both strength and functional testing outcomes.^9,15,17  Power generation in the control group demonstrated no between-limbs differences, while the ACLR group demonstrated a higher percentage of power at the hip in the operated limb than in the nonoperated limb, which showed a similar distribution as in the healthy limbs. This finding, further compounded by the identification of hip-joint power as the strongest predictor of group membership, highlights the importance of examining jumping strategy and jumping height when assessing rehabilitation status.

Although these results indicate that limb symmetry is important, absolute values in the ACLR limbs did not correlate with those in our control group. We could postulate that nonoperated limbs lose strength and power postoperatively,²⁶  yet in these participants, the nonoperated limb demonstrated strength and power values similar to those in the control group.

The normative values identified in this study add to the literature and offer relative measures that athletes can target as markers of normal function. Normative values can be used as return-to-play criteria, whereby clinicians can be assured that ACLR patients have achieved strength and power that are normal for their age, sex, and activity level, even if the limb strength and performance outcomes are symmetric.

Our study was limited to active males between 24 and 26 years of age. Furthermore, the number of control participants was small. Future authors should aim to establish more generalizable normative values for commonly used outcome measures in ACLR rehabilitation specific to each sex and activity level.

It was also beyond the scope of this study to detail the individual rehabilitation course of each participant, which may be a confounding variable for both the ACLR and control group results. Although all in the ACLR group had undergone surgery at least 6 months earlier, their recovery experiences may have been different. We did not account for how long each participant had been allowed to perform jumping activities. Those who had been cleared more recently may have had poorer jump heights than individuals who had been cleared earlier and had more time to practice and perform this task.

Specific variations in movement patterns between sexes during the early landing phase after a jump have been well established^40,41 ; thus, we deemed it important to include only male participants. The current study results should therefore not be generalized to female athletes.

Six months postsurgery was chosen as an appropriate time for assessment in this study, when subjective knee function is well below normative values. This can be supported by an average score of 68.3 on the International Knee Documentation Committee form (95% CI = 66.6, 70.1) for the ACLR group. For a normative population of men aged 18 to 24 years with no history of knee problems, a score of 89 ± 18 according to the International Knee Documentation Committee has been established.⁴²  This suggests that the rehabilitation within the ACLR group in this study was not complete at the 6-month postoperative testing and is an important consideration when interpreting our results.

We performed multiple comparisons, and one could argue that we should have implemented a multiple-comparisons correction to reduce the type 1 error. However, as the type I error decreases, the chance of type II errors increases.^43–46  Our conclusions were based on P values in combination with effect sizes, and differences with weak effects were handled with care. The reader should note that some significant differences with weak effect sizes would not have been different if we had applied a correction (ie, ISO knee-extension peak torque in the control group and ISO knee-flexion peak torque, SL CMJ peak power LSI, and proportion of power generated at the knee between the ACLR and control groups).

Considering both SL CMJ and ISO strength measures is important when assessing a patient's rehabilitation status after ACLR. The normative and ACLR cohorts differed strongly in absolute and symmetry measures but also in jumping strategies, with increased hip power compensating on the reconstructed side. This was supported by the results of a repetitive regression analysis that determined group membership with 89% accuracy by selecting absolute and symmetry measures of SL CMJ and ISO. As such, measuring absolute strength and performance during both a multijoint closed chain activity as well as ISO strength is beneficial in identifying the athlete's needs during rehabilitation.

Future authors should examine the biomechanical differences between limbs and compare ACLR and healthy participants to identify relevant variables and look prospectively at the influence of strength and power measures on return to play and the risk of a second ACL injury.

This study was funded in part by the Sports Surgery Clinic and by Science Foundation Ireland under the grant number SFI/12/RC/2289.