Investigate the kinematic variables that are related to EPKAM waveforms, and therefore potentially ACL injuries.
The study included 163 athletes (107 female) in soccer or team handball, median age 16 (IQR = 15-17). Athletes were recruited between the years of 2016-2019. In 2024, they were asked for a second consent to find medical record of ACL reconstruction surgery.
Retro-reflective markers were placed on subjects and sampled at 400Hz and a resolution of 0.3MP. Ground reaction forces (GRF) were captured with a force plate capturing at 2000Hz. The athletes performed 10 repetitions per leg of a change of direction task from a ready position and the kinematics and kinetics calculated.
KAM waveforms were transformed to signed differences, and a Ward D2 clustering algorithm was used to categorize them into either the EPKAM waveform, or other waveforms. Kinematics and vertical GRF were independent variables and EPKAM the depdenent variable. A boosted regression model was calculated from 80% of subjects with an interaction depth of 3, shrinkage of 0.01, and a maximum of 7000 trees. The model was cross validated on the remaining 20% of subjects.
Each variable was then excluded sequentially from the model, and when the exclusion resulted in a significant drop of explanatory power that variable was deemed important. A final model with only the important variables was used to calculate the predicted EPKAM for each trial. Mixed logistic regression models were calculated to compare the observed and predicted EPKAM as risk factors for ACL injuries, and compared against a base mixed model.
Seven athletes had suffered ACL injuries, all female. The only variables that could meaningfully predict the EPKAM were the knee adduction angle 10°, and hip abduction angle, both at initial contact. The correlation between predicted and observed EP was 0.41 (from cross validation of 20% of subjects not included in boosted regression). The base mixed model had an AIC of 181.49. The predicted EPKAM (AIC = 157.37) was a stronger risk factor for ACL injuries than the observed EPKAM (AIC = 168.18).
The interaction between hip and knee abduction angle at initial contact is likely important in terms of the EPKAM waveform. Two important break-points were discovered. Between 10° of knee adduction and 10° of knee abduction, the EPKAM waveform was increased with an interaction between knee and hip abduction angles. At more than 10° knee abduction, however, there was always a high chance of the EPKAM waveform.
Movement strategies involving knee and hip abduction are likely to produce EPKAM waveforms, and are risk factors for ACL injuries but with two non-linear break-points. Regular linear statistics may be insufficient to explore potential risk factors for ACL injuries due to the presence of non-linear break points in the relationships.
movement patterns
acl injury prevention
