Institutional members access full text with Ovid®

Development and Validation of a Clinic-Based ACL Injury Risk Prediction Algorithm for Use in Targeted Neuromuscular Injury Prevention Training.

Myer, G D; Ford, K R; Khoury, J; Succop, P; Hewett, T E
Journal of Strength & Conditioning Research: March 2011
doi: 10.1097/01.JSC.0000395584.33628.a8
Abstract: PDF Only

Prospective measures of high knee abduction moment (KAM) during landing identify female athletes at increased risk for ACL injury. However, dedicated laboratory space, costly measurement tools and labor intensive data collection sessions are necessary for these measures. PURPOSE: To develop and validate clinic-based estimates of KAM in an ACL injury risk prediction algorithm. METHODS: Female basketball and soccer players (N = 744) were recruited for testing of anthropometrics, maturation, laxity/flexibility, strength and landing biomechanics. Linear regression was used to model KAM and logistic regression was used to examine high versus low KAM as surrogate for ACL injury risk. For validation purposes, 20 female basketball, soccer and volleyball players were tested using 3-dimensional (3-D) motion analysis and clinic-based techniques simultaneously. Clinic-based measurements were validated against 3D motion analysis measures using within and between method reliability (intraclass correlations (ICC) and Bland-Altman Plots) and sensitivity and specificity comparisons. RESULTS: The most parsimonious linear regression (R2 = 0.78) model included the independent predictors ([beta] +/- 1SE): 1) peak knee abduction angle (1.78 +/- 0.05; P<0.001), 2) peak knee extensor moment (0.17 +/- 0.01; P<0.001), 3) knee flexion ROM (0.15 +/- 0.03; P<0.01), 4) body mass index Z-score (-1.67 +/- 0.36; P<0.001) and 5) tibia length (-0.50 +/- 0.14; P<0.001). The logistic regression model that employed these same variables predicted high KAM status with 85% sensitivity and 93% specificity and a C-statistic of 0.96. Clinical correlates to laboratory-based measures were identified and predicted high KAM status with 73% sensitivity and 70% specificity. The clinic-based prediction algorithm, including (Odds Ratio [OR], 95% confidence interval [CI]) knee valgus motion (cm) (OR = 1.43, 95% CI:1.30-1.59), knee flexion ROM (deg) (OR = .98, 95% CI: 0.96-1.01), body mass (kg) (OR = 1.04, 95% CI:1.02-1.06), tibia length (cm) (OR = 1.38, 95% CI: 1.25-1.52) and quadriceps to hamstrings ratio (%) (OR = 1.70, 95% CI: 1.06-2.70) predicted high KAM status with C statistic 0.81. In the validation dataset, the within variable and between analysis showed good reliability with ICCs that ranged from moderate to high, 0.60 to 0.99. Bland-Altman plots confirmed that each variable provided no systematic shift between 3D motion analysis and clinic-based methods and no association between the difference and average. Regression analysis validated previous models with 80% and 75% prediction accuracy for 3D motion analysis and clinic-based techniques respectively. Conclusion: Clinically obtainable correlates derived from highly predictive 3D motion analysis models demonstrated high accuracy for determination of high KAM status in the ACL injury risk prediction algorithm. PRACTICAL APPLICATIONS: The current development and validation steps provide the critical next progression to bridge the gap between laboratory identification of injury risk factors and clinical practice. Implementation of the developed prediction tool will likely increase both the efficacy and efficiency of prevention strategies for non-contact ACL injury and its widespread use may impact the geometric rise of this injury in female athletes. ACKNOWLEDGMENT: Funding support from National Institutes of Health/NIAMS Grants R01-AR049735, R01-AR05563 and R01-AR056259.

(C) 2011 National Strength and Conditioning Association