LINEAR PREDICTION OF KINEMATIC AND KINETIC PERFORMANCE WITH INDIVIDUAL ANATOMIC FACTORS Nathaniel A. Bates1, Greg D. Myer1,2, Timothy E. Hewett1,2,3 1
University of Cincinnati, Cincinnati, OH, USA Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, USA 3 OSU Sports Medicine Sports Health & Performance Institute, Ohio State University, Columbus, OH, USA email:
[email protected] website: http://cincinnatichildrens.org/sportsmed 2
INTRODUCTION Anterior cruciate ligament (ACL) ruptures are expensive, debilitating athletic injuries that lead to early onset osteoarthritis. [1] Kinematic and kinetic variables observed in tasks such as the drop vertical jump (DVJ) have been associated with increased ACL injury risk. [2] In order to prevent ACL rupture it is important to identify athletes that exhibit injury risk biomechanics, but the cost of motion capture systems needed for full biomechanical analysis is prohibitive to many institutions. As such complex nomograms have been developed to predict an athlete's injury risk biomechanics sans motion capture. [3] Early identification of injury risk mechanics may lead to injury prevention through the application of prophylactic interventions. [4] However, in some situations, such as the study of motion on cadaveric specimens, an investigator may lack some of the data points necessary to calculate a nomogram. The reproduction of in vivo motion on cadaveric specimens allows investigators to study injury mechanics in novel ways. Therefore, it is worthwhile to investigate how effective individual factors are at linearly modeling kinematic and kinetic performance. The purpose of this investigation was to establish whether individual anatomical factors can accurately predict kinematic and kinetic performance values during a DVJ task. METHODS 239 female basketball players at the middle and high school level were evaluated through a series of biomechanical tests. Each subject was assessed for height with a stadiometer and weight with a calibrated physician’s scale. Subjects were instrumented with 43 retroreflective markers and
asked to perform three DVJ tasks from an initial height of 31 cm. 3D motion data was collected during each DVJ task with a 10-camera Motion Analysis Corp. system and dual AMTI force platforms. Kinematic and kinetic data were then calculated in Visual3d with a subject-specific biomechanical model. Isokinetic knee extension, knee flexion, and hip abduction strength were collected with a BIODEX dynamometer at a revolution speed of 300º /sec for the knee and 120º /sec for the hip. From this data four anatomical factors were selected to be evaluated against 12 kinematic and 12 kinetic measures for linear correlations. MATLAB was used to calculate the significance of each linear regression model and corresponding R-squared values. Statistical significance was determined at α < 0.05. RESULTS AND DISCUSSION Most linear models exhibited poor R-squared values (Table 1). Anatomical factors expressed significant relationships (P < 0.05) with kinematic measures in 26 of 48 cases. However, R-squared values were universally poor (R2 < 0.18). Similarly, though significant relationships (P < 0.05) were identified in 39 of 48 cases for kinetic measures, the maximum R-squared value was 0.69. Poor Rsquared values indicate that, while significant, most of these models did not account for enough error to be clinically applicable predictors. The exceptions to this were height and mass relative to peak adduction moment (Figure 1) and flexion moment range. Unfortunately, these two measures have not been identified as indicators of ACL injury risk. Combination of factors, such as the height*mass, did not increase the number of measures for which models were significance or improve R-squared values relative to individual factors.
Table 1: Displays correlations between anatomical factors (columns) and performance measures (rows). Highlighted cells indicate a significant relationship was present and the corresponding Rsquared value is indicated within that cell. Ht Mass H*M Tib Kinematics Flexion @ initial contact 0.03 0.02 Abduction @ initial contact 0.06 0.05 Internal @ initial contact 0.04 0.04 0.05 0.05 Max flexion angle 0.12 0.17 0.18 0.02 Min flexion angle 0.06 0.05 Max abduction angle 0.04 0.03 Max adduction angle 0.06 0.08 0.08 0.02 Max internal angle 0.09 0.08 Max external angle 0.04 0.04 Flexion ROM Abduction ROM 0.06 0.06 Internal ROM Kinetics Flexion @ initial contact 0.02 0.07 0.06 Abduction @ initial contact Internal @ initial contact Max Flexion Moment 0.12 0.32 0.31 0.04 Max Extension Moment 0.07 0.17 0.17 0.05 Max Abduction Moment 0.10 0.17 0.18 0.05 Max Adduction Moment 0.42 0.63 0.65 0.23 Max Internal Moment 0.09 0.07 0.08 0.06 Max External Moment 0.08 0.05 0.06 0.05 Flexion Moment Range 0.42 0.68 0.69 0.22 Abduction Moment Range 0.27 0.38 0.39 0.19 Internal Moment Range 0.25 0.27 0.30 0.14 Ht = subject height; Mass = subject mass; H*M = subject height multiplied by subject mass; Tib = subject tibia length
More complex models are necessary to correlate factors with most kinematic and kinetic measures. The nomogram developed by Myer et al. [3] predicts a single kinetic measure based on a multitude of anatomic, strength, and kinematic factors. Most of the current models are too oversimplified to accurately predict performance. As these linear prediction models do not accurately estimate kinematics, the adjustment of input kinematics relative to specimen size in cadaveric studies is not justified. Similar to clinical prediction, the poor R-squared values from linear models applied in this study account for too little kinematic variability to justify change. Therefore in studies that aim to recreate in vivo motion in cadaveric specimens, alteration of kinematic inputs relative to specimen size would be inappropriate. Significant correlation between anatomic factors and kinetic measures were expected. Mechanics dictate that increased mass at a further distance will generate greater moments. However, even these factors mostly did not account for enough variation to be predictive beyond broad generalizations. CONCLUSIONS Linear models that correlate anatomical factors to kinematic and kinetic performance may not account for sufficient variance to accurately predict values. More robust, multi-factor nomograms are necessary to predict values and clinically significant variables such as relative ACL injury risk. [3] REFERENCES 1. Delince P, et al. Knee Surg., Sports Traumatology, Arthroscopy. 20, 48-61. 2009. 2. Hewett TE, et al. Am. J. Sports Med. 33, 492501. 2005. 3. Myer GD, et al. Am. J. Sports Med. 38, 20252033. 2010. 4. Myer GD, et al. J. Strength Cond. Res. 20, 345353. 2006. ACKNOWLEDGEMENTS
Figure 1: Graphic depiction of kinetic performance against anatomical factor (height*mass) data.
Funding supported by NIH Grants R01-AR049735, R01-AR055563, and R01-AR056259.