Backward elimination multivariate analysis exploring variables associated with oxygen and carbon dioxide transfer during ECMO support.

This multivariate analysis was performed using a generalized linear mixed model adjusted to each animal with a backward elimination.The initial dependent variables in the oxygen transfer analysis were blood flow, O2 saturation, PaCO2, temperature, base excess, pH, sweep flow and hemoglobin. Base excess and pH were eliminated from the analysis in order to reduce the VIFs bellow the threshold of 2.5, keeping the PaCO2 in the model. Sweep flow, temperature and hemoglobin dropped out during the backward elimination of the multivariate analysis. The coefficient of determination of the final model (pseudo - R2) was 0.81.The initial dependent variables in the carbon dioxide transfer analysis were blood flow, O2 saturation, PaCO2, temperature, sweep flow and hemoglobin. Temperature and O2 saturation dropped out during the backward elimination of the multivariate analysis. The coefficient of determination of the final model (pseudo - R2) was 0.89.§The blood samples were acquired from the pre-membrane port.*VIF denotes variance inflation factor.#Beta-unstandardized coefficient denotes the estimated variation of the oxygen transfer in mL/min for each unit (the units are cited in the table) variation of the independent variables.