Multiple regression analysis, using body size and cardiac cycle length, in predicting echocardiographic variables in dogs.

A significant (P less than 0.0001) positive correlation was demonstrated between left ventricular internal chamber dimension in diastole or systole and body weight, body surface area, cycle length, and the square root of cycle length. On the basis of adjusted coefficients of determination, multiple regression analysis, using body weight or body surface area and cycle length or the square root of cycle length, was superior to separate simple regression with these variables in accounting for variations in left ventricular internal chamber dimensions. Shortening fraction had a significant (P less than 0.0001) negative correlation and left ventricular free wall measurements had a significant (P less than 0.0001) positive correlation to body weight and body surface area. For these echocardiographic variables, correlation to the square root of cycle length was insignificant (P greater than 0.05), and a multiple regression model was not helpful in developing confidence intervals. Septal wall measurements were not correlated with body weight, body surface area, cycle length, or the square root of cycle length.