Cardiac output assessment methods in left ventricular assist device patients: A problem of heteroscedasticity

Abstract

Equipoise remains about how best to measure cardiac output (CO) in patients with left ventricular assist devices (LVAD). In this study, direct Fick CO was compared with thermodilution (TD) and indirect Fick (iFick) CO in 61 LVAD patients. TD and LaFarge iFick showed moderate correlation with direct Fick (R2=0.49 and R2=0.38, p < 0.001 for both), while Dehmer and Bergstra iFick showed poor correlation with direct Fick (R2=0.29 and R2=0.31, p < 0.001 for both). Absolute bias between all CO estimation techniques and direct Fick CO was lowest for TD compared to iFick methods but significant for all methods. All methods tended to overestimate CO compared to direct Fick, with greatest overestimation present in those with the lowest measured direct Fick CO. Bias and frequency of significant discrepancy were least using TD and Lafarge iFick CO estimation methods in this study, with TD CO demonstrating modestly better correlation and less heteroscedasticity compared to Lafarge.