Motor learning strategies in basketball players and its implications for ACL injury prevention: a randomized controlled trial

Informed written consent was obtained prior to inclusion. First, anthropometric measures were taken prior to placement of 21 reflective markers of 14 mm in diameter placed according to the Vicon Plug-in-Gait marker set, with additional trunk markers on the sternum, clavicle, C7, T10 and right scapula. This was followed by a static calibration. All subjects wore spandex shorts and shirts (for females) and their own athletic shoes. The tasks chosen for this experiment were a 45° sidestep cut, straight run or 45° crosscut, which were randomly indicated with a green light. The straight run and 45° crosscut were used as additional tasks in order to present the subject with three options (sidestep cut, run and crosscut) [3, 41, 42]. Subjects used a 5-m approach run followed by a 1-foot landing on the force plate and a 45° change in direction through a second set of timing gates 5 m away from the force plates. To reassure standardization while challenging athletes, both approach and exit speed had to be within 4.5–5.5 m/s. Subjects were instructed to land on the force plate with their dominant leg, which was defined as the leg they prefer pushing off with while jumping. The general instruction for all groups was: ‘run towards the force plate and just before the force plate you will see a bright green light showing up, indicating the direction you have to run after placing your foot on the force plate. After passing the forces plate continue running until you have passed the timing gates’. Each subject was given ample time to warm up and familiarize themselves with the set-up.

After familiarization, the first three trials served as a baseline. After baseline collection, the first session (S1) was started, in which feedback was provided to the VIS and VER groups after every correct trial. In S1, the VIS group received video feedback on a TV screen (LG, Flatron 65VS10-BAA) showing the subject from behind. A new video was only shown if the subject performed better than the previous best trial; this became the new best trial (i.e. smaller external peak knee frontal plane moment). No explicit feedback or instructions were given; however, subjects knew that they were looking at their best trial so far. They were instructed to replicate that trial to the best of their ability. In SI, the VER group received the following instructions: (1) ‘bend your trunk forward’, (2) ‘bend your knee’ and (3) ‘keep your knee straight above your foot’. The CTRL group was only provided with the general instructions. Two retention sessions were conducted, after 1 week (S2) and 4 weeks (S3). No feedback at all was given during the retention tests.

Thirty-five successful sidestep cutting trials, defined as correct speed and cut angle of 45° (marked with tape) with the correct foot on force plate, were collected per session. Subjects were aware of the location of the force plate, but as they were focusing on the light stimuli, targeting of the plate was avoided. Also, tape was placed at the start of the approach distance to facilitate the desired foot contacting the force plate. Trials were rejected if the subject clearly targeted the plate (i.e. a ‘stutter step’ or ‘reaching’). Each subject was given enough rest between trials to reduce the potential effects of fatigue.

Kinematic data were collected using an 8 camera motion analysis system at 200 Hz [Vicon Motion Analysis Systems Inc., Oxford, UK and Vicon Nexus software (version 1.8.3, Oxford, UK)]. Good measurement accuracy and high test and retest repeatability have been previously reported [22, 31]. Ground reaction force data were collected at 1000 Hz with two Bertec force plates (Bertec Corporation, Columbus, OH). To provide feedback to the VIS group, a Basler camera (640 × 480, 210 fps, Vicon Motion Systems, Inc., Centennial, CO) with a 25-mm C-mount lens was used to collect analogue high-speed data. Two infrared timing gates (HL 2-31 Photocell, TAG Heuer professional timing, Switzerland) were used to ensure that running speed was 4.5–5.5 m/s. A 3-light guiding system was used to randomly cue the subject 0.5 s before stepping on the force plate [34].

Procedures were approved by the University of Groningen Medical Ethics Committee (ID numbers: CCMO protocol number: NL24814.042.09, METc: 2009.142).

Based on previous research [34, 41, 48], sample size was estimated for a minimal statistical power of 80 % (α = 0.05). All sample size and power calculations were completed using (G*Power for Mac, version 3.1.2). Given the variation of the dependent measures that were included in this study, 15 subjects per group (male and female) were deemed adequate. Hence, 15 females and 15 males were allocated to the VIS, VER and CTRL groups, respectively.

Primary outcome variables were vertical ground reaction force (vGRF), sagittal angles and moments of the trunk, hip, knee, ankle and range of motion (RoM). For the knee, frontal plane moments were also collected. All variables are expressed at peak external valgus/varus moment. RoM was calculated as the value at peak external valgus/varus moment minus the value at initial contact. Moments are expressed as external moments normalized to body weight. Results in degrees will be reported to one decimal case [53]. Based on number of subjects and standard deviation, effect sizes (ESs) were calculated for all comparisons. Cohen’s d values are reported as a measure of ES, where 0.2 ≤ d ≤ 0.5, 0.5 ≤ d ≤ 0.8 and d ≥ 0.8 represent a small, moderate and large effect, respectively [11].

Sidestep trials were analysed of all included subjects (n = 30 VIS, n = 30 VER, n = 30 CTRL). Customized software using MATLAB 6.1 (The MathWorks Inc., Natick, MA) was written and used to compute segmental kinematics and kinetics of the tested leg. Force plate and kinetic data were filtered using a fourth-order zero-lag Butterworth low-pass filter at 10 Hz. Assumptions for normality of distribution for all variables were checked. Assumptions of homogeneity of variance and sphericity were also validated for the use of analysis of variance (ANOVA). Differences between groups at baseline were determined using a multivariate ANOVA. To determine differences between groups (VIS, VER, CTRL), time (S1, S2, S3) and sex (female, male), a 3 × 3 × 2 repeated measures ANOVA was conducted followed by post hoc comparisons (Bonferroni) with alpha level set at α ≤ 0.05 a priori. Additionally, the start and end values of each session were calculated (linear fit of all data).

Overall, the major changes were observed within the male VIS group: they acquired better motor skills (S1) with retention (S2, S3). Their adopted landing strategy included larger vGRF [8.5 % increase in S1, p = 0.042 (Fig. 1)], knee flexion moment, knee RoM and ankle dorsiflexion angle. The soft landing strategy with larger knee RoM and ankle dorsiflexion angle allows for better load dissemination [12]. As the knee flexion angles were greater than 50°, along with a reduction in the knee valgus moment over time, this likely reduces combined loading of the ACL and the risk of ACL injury. Males in the VIS group seemed to be effective in performing the task by flexing their knee and absorbing mechanical energy while actively aligning the vGRF close to the centre of the knee joint (Fig. 2a). As the increased vGRF only had an impact on the sagittal knee moments, we feel this is not a major concern [33] and a reflection of increased motor control during this task. During heavy deceleration and a short contact time, there is no increase in frontal plane moment and the males use their quadriceps to execute the sidestep with a flexed knee to forcefully push off the force plate.

These findings are in agreement with others examining the effect of IF and EF instructions in healthy athletes, where the EF group increased knee flexion RoM during a counter movement jump [29]. Others have found a reduction in vGRF in double-legged jumps following video feedback [40, Dallinga et al. 2015, manuscript in review]; however, the task, sex, different modes or combinations of feedback may all contribute to the differences found. A direct comparison therefore cannot be made.

Frontal plane moments were less responsive to the VIS or VER feedback provided. In contrast, Dempsey et al. [14] noted reduced knee valgus moment following video feedback. These athletes saw both expert and own performances and received individual verbal feedback based upon required changes [14], and this combination of visual (expert and self) and verbal feedback could have positively influenced their results [43]. In double-legged jumps, knee valgus displacement was also positively influenced in two studies using visual feedback [17, 35], while no changes were found in another study [51]. Also providing subject views from different sides (front, side, back) may have resulted in better adoption of the required technique. With a complex manoeuvre such as sidestep cutting, more comprehensive feedback may be advantageous to achieve whole-body technique modification as knee load is also dependent on trunk control [20]. The instructions and feedback should be kept relatively simple though, as high complexity of feedback hampers motor learning [30].

Of the two feedback techniques (VIS and VER), sex differences were only observed within the VIS group, with males showing larger knee flexion moments and vGRF compared to the females (Fig. 2). A potential explanation is that females use a ‘ligament dominant’ landing [36], which may place them at greater risk of ACL injury [26]. In addition, with reduced neuromuscular control, the direction of the vGRF is less controlled in all planes (Fig. 2b). The females in the VIS group did, however, increase trunk flexion angle over time, which is in favour for reducing potential ACL injury risk [19, 46].

It is not clear why the females in the current study were less responsive to feedback. Learning new strategies can be hampered in females if they use their muscles less effectively. Maybe females need more time to adopt a safe landing strategy. Additionally, the females in our study can be classified as ‘low-risk’ females (based on knee valgus moment), who have been shown to have smaller potential to change their movement pattern compared to ‘high-risk’ females [37]. Moreover, it is plausible that females in general prefer different learning strategies (e.g. combination of visual and verbal feedback) [24], considering the lack of effect in the females in the VIS group and the stronger effect in the female VER group for some of the variables (Fig. 1). Adding verbal instructions could enhance motor learning in females. However, literature shows that verbal instructions that induce an EF yield better results in motor skill acquisition and retention compared to IF instructions [54]. In another study of this research group, subjects performed a drop vertical jump with self-controlled expert video feedback, demonstrating beneficial results of verbal EF feedback in females and video feedback in females and males on landing technique, with one week retention (Welling et al. 2015, manuscript in review).

One group received instructions (VER, no change in wording), while the other group received feedback (VIS, change in video). This disparity could have affected motivation and therefore learning effect. Experiencing competence through good performance (knowing that you are watching your best trial, being involved in analysing own performance) adds to the evidence of motivational influences on motor learning (e.g. intrinsic motivation, interest and enjoyment, positive comparative feedback) [10]. Also, we did not examine how exactly subjects adhered to the instructions provided. Findings of this study may be limited to this specific population of recreational athletes and may not translate to other populations such as elite or younger athletes. Furthermore, the accuracy of skin-based markers in estimating joint kinematics and kinetics has been questioned [44].

Lastly, when considering the benchmarks from Cohen referred to in this manuscript, especially for the kinetic variables, most effect sizes are large. However, this is caused by the small standard deviations. Our group was homogeneous, and data collection was very stable and therefore reliable. The differences of the means do have a clinical relevance as indicated in the discussion, i.e. the changes depicted in the males receiving visual feedback reflect a potential safer landing strategy.

More evidence is needed regarding biomechanical risk factors for ACL injury in male athletes [2]. Future research should also examine the use of combinations of self-feedback (reviewing own video) and expert modelling (reviewing video from an expert), along with providing feedback ‘on demand’ (self-controlled learning) with verbal and visual cues. Future research should also focus on transfer to the field to examine whether the learned movement techniques remain during a practice or game. It is hypothesized that learned skills with an EF transfer better to the field [6].

Sex-specific learning preferences should be acknowledged when implementing ACL injury prevention programmes as males responded well to visual feedback, while females did not. For injury prevention, a safer landing technique as well as stable or increased performance is crucial [23]. The results of the current study and others [14] support the use of EF feedback. ACL injury risk may be reduced in males by an increase in knee flexion moment and decrease in knee valgus moment, respectively, whereas performance (running speed) was not compromised. These feedback techniques seem to be advantageous in contrary to IF verbal instructions leading to a decrease in performance (i.e. decreased jump height or movement speed) [13, 28, 49, 52].

The goal of ACL injury prevention is to achieve long-term effects and transfer learned skills to actual competition. None of the previous mentioned studies measured retention to examine whether a permanent change was achieved. In our study, retention was achieved after one and 4 weeks in the males in the VIS group as they continued to demonstrate superior technique. This indicates that the effect was not only immediate and temporary but also relatively permanent. Apparently, the learning process initiated with the visual feedback continues by repeating motor patterns in the brain even if no feedback is given (motor imagery) [55]. This is a very interesting given as this implicates that learning with a visual component stimulates automaticity and retention and therefore might require less time and investment from training staff. Recently, a single 15-min video feedback protocol led to greater reduction in dynamic knee valgus during a drop vertical jump task [35] than that has been observed after a four-week jump training programme [18].

The ‘whole-body approach’, that enhances being embedded in the task (embodied cognition), provided by visual feedback appears to be an effective method to promote motor learning [14]. Imitation plays an important role when receiving visual input [6]. The mirror neurons facilitate motor learning by automatically mapping observed movements onto a motor programme without high cognitive involvement [45].

Overall, the results of the present study demonstrate that learning through observation and practice is a powerful tool. The current study expands on the previous published conceptual and practical framework to enhance effectiveness of current ACL injury prevention programmes [4, 6]. Cumulatively, evidence is emerging that implementation of EF feedback especially with a visual feedback component is promising in terms of reducing ACL injury risk while maintaining or increasing performance.

In conclusion, sex-specific learning preferences may have to be acknowledged in day by day practice. Adding video instruction or feedback to regular training regimens when teaching athletes safe movement patterns and providing individual feedback might target suboptimal long-term results and optimize ACL injury prevention programmes.