Abstract
Introduction: There are multiple modalities used to measure cardiac output in pregnancy. The accepted gold standard is cardiac catheterisation, then measurement via direct Fick method or thermo-dilution. These methods are invasive and not feasible for assessment of healthy pregnant women. There are several non-invasive devices widely used to measure cardiac output including Innocor, an inert gas re-breathing technique which has been validated against thermodilution [1–3] and direct Fick methods [4,5] for measurement of pulmonary blood flow and thus, cardiac output. Measurements of cardiac output by non-invasive cuff-based device (Vicorder) has been validated against Innocor [6], however, this was in a health non-pregnant cohort, and validation between these two devices has not been previously reported in a pregnancy cohort.
Methods: Healthy women who were recruited as part of a preconception cohort undergoing comprehensive cardiovascular measurements were seen at 6, 10, 22 and 34 weeks of pregnancy once they had conceived. In total, there were 40 women seen at each gestation band, giving a total of 160 measurements. For each, cardiac output was measured in the left lateral lying position using the Vicorder and Innocor device. All the measurements were performed by one assessor. All statistical analyses were carried out using SPSS and MedCalc. Agreement between both methods of cardiac output estimation was assessed using Bland Altman plots and statistics (Mean cardiac output, bias (mean difference), limits of agreement (bias ± 1.96 SD of bias) and percentage error ((1.96 SD of bias/mean CO) X 100%) and intra-class correlation (ICC). A low bias indicating accuracy, narrow limits of agreement with percentage error <30% reflecting precision and ICC 40.7 are considered to reflect good agreement between two methods of CO determination.
Results: Bias between the two methods was –0.12 l/min. 95% limits of agreement were between +2.5184 and –2.7716. Percentage error was 41%. ICC between both methods of assessment was 0.597.
Conclusions: While bias is low, percentage error and ICC indicate only moderate agreement between Innocor and Vicorder for CO measurements.