Precision and accuracy of ground reaction force normalization in a heterogeneous population of dogs
Ursula Krotscheck1, Rory J Todhunter, Samantha A Nelson
1Department of Clinical Sciences, Cornell University, Ithaca, New York.
Objective:
To determine if currently used ground reaction force (GRF) normalization methods are accurate and precise enough to be used on a single-limb basis.
Study Design:
Prospective clinical trial.
Animals:
Clinically normal (n = 69) dogs and 40 dogs with unilateral ruptured cranial cruciate ligaments (CCL).
Methods:
Pelvic limb GRFs of orthopedically normal dogs and those with unilateral ruptured CCL were collected. Normalization methods included none, body weight (BW), withers height (WH), WH and relative velocity (WH*F) and principal component 1 (PC1). Normalization methods were evaluated both by individual GRFs and additively. Binary logistic regression was performed for all normalization methods; sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) calculated. Stepwise backward logistic regression was used; significant values were retained in the final model. P < .05 was significant.
Results:
Normalization of contact time (CT) by BW uniformly increased sensitivity, specificity, PPV, NPV, and accuracy. SI was the most accurate at both the walk and trot (accuracy 80-96%). Normalization by BW, WH, and WH*F all achieved similar results. When normalized GRFs were added, the accuracy increased only at the walk.
Clinical Significance:
CT should be normalized to BW. SIs remain the gold standard, if SIs cannot be used, combining GRFs normalized to BW will result in high precision (80%) and high accuracy (89.5%) at the walk. At the trot, normalization by BW, WH and WH*F results in consistent results for the individual GRFs, though not all accuracies are >80%.
