NONINVASIVE MEASUREMENT OF CARDIAC-OUTPUT BY THORACIC ELECTRICAL BIOIMPEDANCE - A STUDY OF REPRODUCIBILITY AND COMPARISON WITH THERMODILUTION

The performance and reproducibility of the BoMED NCCOM3 thoracic electrical bioimpedance cardiograph (TEB) has been evaluated in volunteers and patients. In resting supine volunteers, we determined the coefficient of variability over short time periods (30 min) and over several days, and examined the effects of differences in electrode type and electrode placement. The mean (range) intra-subject coefficients of variation (CV) for thoracic fluid index (TFI) and stroke volume (SV) were 1.0% (0.4-1.8%) and 4.7% (2.1-8.5%), respectively over a 30-min period. The corresponding CV were 5.6% (2.3-10.9%) and 10.9% (6.1-14.8%) for measurements made at rest on four separate occasions. Use of different electrode types (RedDot and Medicotest) resulted in differences in TFI (P < 0.01), but not in mean values for SV or cardiac output (Q); their use in individual subjects revealed differences of up to 20% in SV and Q. Alterations in electrode placement by 5 cm in the horizontal and diagonal planes produced no significant changes in TFI, SV or Q; changes in the longitudinal plane produced a graded change. Increases of 5 cm and 10 cm in thoracic length produced mean increases in TFI of 9.8% and 39.8%, respectively, and mean decreases in Q of 8.4% and 16.7% and SV of 7.5% and 15.8%. TEB measurements of Q and SV were compared with thermodilution (TD) in 16 intensive care patients. Mean (SEM) Q by TEB was 5.63 (1.10) litre min-1 compared with TD 4.38 (0.72) litre min-1 (P < 0.01). The correlation coefficient for Q (r) was 0.72, producing the equation Q(TEB) = 0.835 x Q(TD) + 1.13; that for SV was r = 0.83, SV(TEB) = 1.009 x SV(TD) + 6.84. Bias between the two methods revealed a mean difference of -0.86 litre min-1 (95% confidence limits -2.6 to +0.89) for Q, and -13 ml (-35.2 to +9.2) for SV. Analysis of individual patient bias plots revealed two different trends. If the range of Q values was < 1.0 litre min-1, there was a variable bias of magnitude similar to that seen in the volunteer studies. If values of Q had a range of > 1.0 litre min-1, there tended towards a constant bias for that individual patient, indicating that TEB is able to follow trends.