Aymen A Alian1, Terence Rafferty. 1. Department of Anesthesiology, Yale University School of Medicine, 333 Cedar St., P.O. Box 208051, New Haven, CT 06520-8051, USA. aymen.alian@yale.edu
Abstract
OBJECTIVE: The objective of this study was to determine the function that best expressed the true shape of the regression line between transgastric short axis (TGSA) transesophageal echocardiographic (TEE)) views of the left ventricle (LV) and radionuclear LVEF. METHODS: The literature was searched for relevant articles published between 1979 and 2007. Articles that directly compared TGSA LVEF with radionuclear LVEF were reviewed. Inclusion criteria included the provision that TGSA estimations be acquired by manual tracing of the endocardial border. Digital tabular data sets were created from electronically scanned graphic data (Photoshop, Adobe Systems, Inc., San Jose, CA). Software consisted of SPSS (SPSS, Inc., Chicago, IL). Analysis was by regression curve fitting with standard parametric and LOWESS (locally weighted scatterplot smoothing) data-driven models. RESULTS: Three articles met our study design criteria. The studies generated a total of 32 patients with 99 data points. These data were pooled and analyzed. The LOWESS regression surface demonstrated non-linearity. The best "goodness of fit" template function was determined by selecting the function with the highest R-squared value. The best fit (R(2) = 0.807) consisted of a polynomial function with a power trajectory. CONCLUSION: Linear regression analysis provides a linear regression function. Thus, analyses are forced to conform to a straight-line relationship, irrespective of the disposition of the data points. By contrast, parametric and nonparametric regression models allow a choice of functions. This feature permits construction of a more appropriate "goodness of fit" curve. In this study, our purpose was to determine the function that best expressed the true shape of the regression line. Results indicated that the relationship between TEE TGSA and radionuclear LVEF was curvilinear.
OBJECTIVE: The objective of this study was to determine the function that best expressed the true shape of the regression line between transgastric short axis (TGSA) transesophageal echocardiographic (TEE)) views of the left ventricle (LV) and radionuclear LVEF. METHODS: The literature was searched for relevant articles published between 1979 and 2007. Articles that directly compared TGSA LVEF with radionuclear LVEF were reviewed. Inclusion criteria included the provision that TGSA estimations be acquired by manual tracing of the endocardial border. Digital tabular data sets were created from electronically scanned graphic data (Photoshop, Adobe Systems, Inc., San Jose, CA). Software consisted of SPSS (SPSS, Inc., Chicago, IL). Analysis was by regression curve fitting with standard parametric and LOWESS (locally weighted scatterplot smoothing) data-driven models. RESULTS: Three articles met our study design criteria. The studies generated a total of 32 patients with 99 data points. These data were pooled and analyzed. The LOWESS regression surface demonstrated non-linearity. The best "goodness of fit" template function was determined by selecting the function with the highest R-squared value. The best fit (R(2) = 0.807) consisted of a polynomial function with a power trajectory. CONCLUSION: Linear regression analysis provides a linear regression function. Thus, analyses are forced to conform to a straight-line relationship, irrespective of the disposition of the data points. By contrast, parametric and nonparametric regression models allow a choice of functions. This feature permits construction of a more appropriate "goodness of fit" curve. In this study, our purpose was to determine the function that best expressed the true shape of the regression line. Results indicated that the relationship between TEE TGSA and radionuclear LVEF was curvilinear.
Authors: J H Urbanowicz; M J Shaaban; N H Cohen; M K Cahalan; E H Botvinick; K Chatterjee; N B Schiller; M W Dae; M A Matthay Journal: Anesthesiology Date: 1990-04 Impact factor: 7.892