Linear, logarithmic, and polynomial models of M-mode echocardiographic measurements in dogs

Objective-To determine whether logarithmic and polynomial models are superior to simple linear models for predicting reference values for M-mode echocardiographic variables in dogs with a wide range of body weights. Animals-69 apparently healthy adult male and female dogs of various breeds, ages (range, 1 to 12 years; median, 3.5 years), and body weights (range, 3.9 to 97.7 kg; median, 25.4 kg). Procedure-Echocardiographic M-mode measurements of the interventricular septum, left ventricular dimension (LVD), left ventricular wall, aorta, and left atrium were obtained. Simple linear, second-order polynomial, third-order polynomial, and logarithmic regression models were determined by use of the least-squares method to describe the relationship between M-mode measurements and body weight. Differences in adjusted R-2 values of logarithmic and polynomial models were tested for significance of contribution, compared with the simple linear model. Results-Significant differences in adjusted R-2 were found when comparing simple linear with logarithmic or polynomial models for LVD-diastole, LVD-systole, aorta, and left atrium. Differences in adjusted R-2 between second-order polynomial, third-order polynomial, and logarithmic models were not significant for any M-mode measurement. Conclusions and Clinical Relevance-In this study, logarithmic or second-order polynomial models predicted reference values of M-mode measurements for size of the cardiac chambers better than simple linear models for dogs with a wide range of body weights. Logarithmic and polynomial models were not superior to simple linear models for M-mode measurements of cardiac wall thickness.