Cross-comparison of the trending accuracy of continuous cardiac output measurement devices in postoperation patients
© Kim et al; licensee BioMed Central Ltd. 2009
Published: 13 March 2009
Arterial pulse analysis estimates of cardiac output (CO) are less invasive than pulmonary artery catheter (PAC)-derived ones. Although PAC bolus thermodilution CO (COtd) values are the reference CO values for most clinical studies, there is no defined gold standard. We therefore compared the covariance of three commercially available arterial pulse analysis devices (FloTrac, LiDCO, PiCCO) and PAC with COtd and continuous CO (CCO).
Seventeen postoperative cardiac surgery patients were studied for the first 4 hours post ICU admission. CO was measured simultaneously by FloTrac, LiDCO and PiCCO using the same arterial waveform. LiDCO and PiCCO were calibrated to the first COtd measured on ICU admission. Values of CO were compared before and after specific therapeutic interventions (volume, vasoactive or inotropic infusions). Absolute values for CO across all devices were compared by linear regression and Bland–Altman analysis. Dynamic changes for CO across all devices were compared by moment analysis.
In 17 patients, 72 paired simultaneous CO measurements were collected. By linear regression analysis, all devices correlated well with each other for absolute CO as r = 0.87 (PAC–LiDCO), 0.70 (PAC–PiCCO), 0.51 (PAC–FloTrac), 0.80 (LiDCO–PiCCO), 0.65 (LiDCO–FloTrac) and 0.50 (PiCCO–FloTrac) with P < 0.001 in all pairs. By Bland–Altman analysis, the mean bias between CO measurements by each paired device was -0.18 l/min (PAC–LiDCO), 0.30 l/min (PAC–PiCCO), -0.43 l/min (PAC–FloTrac), 0.12 l/min (LiDCO–PiCCO), -0.63 l/min (LiDCO–FloTrac) and -0.73 l/min (PiCCO–FloTrac), with precision (1.96 SD, 95% CI) of ± 1.56, ± 2.26, ± 3.37, ± 1.98, ± 2.97 and ± 3.45 l/min, respectively. By Pearson moment analysis, the dynamic change for CO was well correlated between PAC–LiDCO, PAC–PiCCO, LiDCO–PiCCO and LiDCO–FloTrac (r = 0.78, 0.56. 0.65 and 0.55, respectively). However, FloTrac showed poor correlation with PAC and PiCCO (r = 0.28 and 0.33, respectively).
In postoperative cardiac surgery patients, the absolute CO of arterial pulse analysis devices showed good correlation with each other and PAC COtd and CCO. However, the dynamic change of CO was less correlated with each other than the absolute CO. Because each arterial pulse analysis device showed different correlations with PAC CO, the CO data from one monitoring system can be used carefully in concert with another to drive resuscitation protocols.
This article is published under license to BioMed Central Ltd.