Since its introduction in 1970 [
Given the importance of the circulation in clinical practice, the frequency of clinical interventions, and the limitations of PAC, there is a need for an improved CO measurement and monitoring method. CO2 partial re-breathing [
The Ultrasonic Cardiac Output Monitor, (USCOM) (USCOM 1A, Uscom Ltd., Sydney, NSW, Australia), is a noninvasive, morphometrically calibrated, continuous wave (CW) Doppler ultrasound device which provides an instantaneous, beat-to-beat measure of right- and left-sided CO. CW Doppler is a widely adopted clinical tool with an accuracy measured by Doppler string phantoms of ±2.3% [
The ultrasonic transit-time flow probe (FP) is considered a true gold standard method for measuring beat-to-beat CO with an accuracy of ±1 to 2% [
This study compared the two clinical methods of CO measurement, PAC and USCOM, with measures from an implanted FP in conscious sheep to evaluate the relative accuracy and sensitivities of the methods across a range of outputs at baseline and during pharmacologic interventions.
The study was approved by the Animal Experimentation Ethics Committee of the Howard Florey Institute. Prior to experiment, 10 adult Merino ewes were anaesthetised with intravenous (i.v.) sodium thiopental (15 mg/kg), and, following intubation, anesthesia was maintained with 1.5–2.0% isoflurane in oxygen. An incision was made above the fourth left rib, the periosteum stripped, and the rib resected. The pericardium was opened, and the ascending aorta cleared for implantation of a FP (20 mm, Transonic Systems Inc., Ithaca, NY, USA). The pericardium, periosteum, muscle, and skin were closed in layers. The flow probe cable was tunnelled superficially and exteriorised near the thoracic spine. Antibiotic prophylaxis (900 mg procaine penicillin, Troy Laboratories, NSW, Australia) was administered for three days after surgery. Postsurgical analgesia was maintained with intramuscular injection of flunixin meglumine (1 mg/kg) (Mavlab, Qld, Australia) at the end of surgery, then four and sixteen hours after surgery. A minimum recovery period of fourteen days was allowed prior to any study.
On the day of the study, a single experienced operator inserted a PAC-continuous cardiac output monitor (CCO) (Baxter Healthcare Corp., Irvine, CA, USA) via the right jugular vein into the pulmonary artery under 2% lignocaine local anesthesia and connected it to a Vigilance Monitor System (Baxter Healthcare Corp., Irvine, CA, USA). With the sheep lying on its right side, both the CO measures from the FP and PAC were captured from the flow meter (Transonics Systems) and the Vigilance, and recorded to computer using a CED micro 1401 interface. FP CO waveforms were acquired at a frequency of 100 Hz per beat. Ultrasound coupling gel was applied to the skin over the acoustic window between the third and fifth left ribs, and the upper thorax was insonated by a single experienced operator using a 3.3 MHz CW probe which was manipulated to optimise the transpulmonary Doppler flow profiles on the USCOM monitor. The USCOM device requires the pulmonary valve diameter (PV) to calculate flow volumes, for example, CO and stroke volume (SV). In human subjects this is determined using a proprietary anthropometric algorithm. An equivalent algorithm is not available for sheep, so for each subject, the USCOM values were calibrated to the FP measurements during a calibration phase prior to experimental measurements, with only post calibration data analysed. The USCOM investigator was blinded to the FP and PAC values, and acquired the Doppler signals and stored the flow profiles to the USCOM hard drive. Each stored screen recorded 6–12 consecutive Doppler profiles, depending on the heart rate (HR). The Doppler profiles were later traced to generate output values for each beat and stored to the USCOM hard drive for later collation and comparison with the corresponding FP and PAC measures.
Contemporaneous CO measurements were made using all methods over a baseline period of up to 40 minutes, during i.v. infusion of the inotrope dobutamine (0.2, 0.4 and 1.0 mg/min for 40–50 min), during the postdobutamine recovery phase (10 min), during infusion of the vasopressor arginine vasopressin (AVP) (15 ng/min for 15–20 min), and during a 5 min post vasopressor phase. The order of dobutamine and AVP infusions was randomised, and a recovery period of 30 minutes between treatments was allowed for CO to return to control levels. As this study compared two beat-to-beat methods, USCOM and FP, against a time averaged method, PAC, 6 to 13 sequential measures of CO by USCOM, and FP were acquired at each comparison point on each subject, depending on heart rate, and averaged to represent the mean CO values at each time point for comparison with PAC values. Further no time to measurement after intervention was less than 4 minutes, thus allowing PAC time to respond to any CO changes.
PAC is an uncalibrated method functioning by detection of temperature gradients across a known distance so was not calibrated. However to address any potential for bias, we performed a post hoc calibration of PAC measures to baseline FP values to generate a FP-calibrated PAC series (cPAC), for comparison with FP measures.
Continuous variables were expressed as mean and standard deviation (SD) and data were analysed using SPSS v.16 (SPSS Inc, Chicago, Ill, USA). Analysis included Bland Altman determination of reliability and reproducibility [
A total of 363 CO measures by FP, 370 measures by USCOM and 293 PAC measures were collected from the 10 adult ewes (
Mean CO by FP, PAC, USCOM and cPAC across all measures was
Summary of comparison of methods for all paired measures and all sheep as absolute values and % values (
Mean (l/min) | Bias (l/min) | LOAs (L/min) | Bias % | Precision % | LOAs % | |
---|---|---|---|---|---|---|
FP versus PAC | −17.2 | 47.0 | ||||
FP versus USCOM | 1.0 | 36.4 | ||||
FP versus cPAC | −0.2 | 54.4 |
Percentage bias and precision for all paired measures for FP versus PAC (−17.2% and 47%), FP versus USCOM (1% and 36.4%), and FP versus cPAC (−0.2% and 54.4%).
Bland Altman plots of FP versus PAC showing bias (−17.2%) and LOAs (−64.2% and 29.8%).
Bland Altman plots of FP versus USCOM showing bias (1%) and LOAs (−35.3% and 37.4%).
Bland Altman plots of FP versus cPAC showing bias (0.2%) and LOAs (−54.7% and 54.2%).
Regression analysis demonstrated a correlation between FP and PAC of
The mean dobutamine induced CO change from baseline measured by FP was 35% (
Mean percentage change of CO from baseline (0%) at each intervention and recovery time-point in all sheep by each method.
% Difference | Baseline | Dobutamine | Post Dob | Vasopressor | Post Vaso |
---|---|---|---|---|---|
FP | 0 | 35.4 | 1.2 | −13.7 | −13.2 |
PAC | 0 | 15.7 | 32.5 | −9.5 | −16.3 |
USCOM | 0 | 39.1 | 5.7 | −5.3 | −4.1 |
cPAC | 0 | 13.2 | 30.0 | −4.0 | −13.1 |
Reliable sensitivity to CO change, ROC AUC > 0.70, was achieved by PAC for changes in excess of −30% but not +40%, with AUC values for ±5% changes being 0.524 and 0.496, respectively, indicating random values (Figures
ROC area under the curve (AUC) values for detection of increased and decreased percentage changes of CO relative to FP where
−40% | −30% | −20% | −15% | −10% | −5% | +5% | +10% | +15% | +20% | +30% | +40% | |
---|---|---|---|---|---|---|---|---|---|---|---|---|
PAC | 0.855 | 0.811 | ||||||||||
USCOM | 0.897 | 0.881 | 0.885 | 0.857 | 0.842 | 0.714 | 0.708 | 0.754 | 0.812 | 0.814 | 0.790 | 0.897 |
0.45 | 0.3 | 0.004 | 0.000 | 0.000 | 0.000 | 0.000 | 0.005 | 0.01 | 0.04 | 0.094 | 0.06 |
ROC curve for 70% certainty of detection of 5% decrease in CO from baseline with PAC in red (
ROC curve for 70% certainty of detection of 5% increase in CO from baseline with PAC in red (
Calibration of the PAC to the FP improved the agreement but not the precision nor sensitivity of the method (Table
The FP calibration of USCOM allowed calculation of sheep PV from a regression equation relating sheep weight (kg), where PV = 0.0132 weight + 1.0329 cm.
Despite no history of validation or efficacy, PAC has been a clinical standard for hemodynamic measurement, diagnosis and monitoring, and applied as a research method and reference standard against which new CO measurement methods have been compared. This study found that PAC was an inaccurate measure of CO and was unreliable for detection of CO changes less than 30–40%. These findings may explain the absence of reported outcomes benefit associated with PAC use and raises questions as to its continued use as a gold standard hemodynamic monitor. USCOM demonstrated equivalence to FP measures across COs from 1.9 to 11.7 L/min, and reliably detected ±5% changes in CO.
Prior comparison studies generally demonstrate USCOM to have acceptable agreement with PAC but good agreement with proven CO measures such as FPs and external cardiac pumps. Prior comparisons of USCOM with PAC in postcardiac surgical critical care patients have reported bias’s of 12%, 18%, and 19% [
The Bland Altman method is the standard statistical method for comparison of two clinical methods [
Measurement of circulatory change is central to the function of hemodynamic monitoring, with detection of 15% changes in CO considered to be clinically desirable [
The current data also demonstrate that PAC under measured the CO change associated with inotropic intervention, a critical and common intervention in the deranged circulation, by 20% (relative 66%) compared with the FP, while USCOM differed by 4% (relative 10%). The differences between measured CO changes associated with vasopressors, where the hemodynamic changes were smaller, were less marked (Table
While experimental measures of accuracy provide insight into the theoretical capability of a modality, clinical utility remains the test of effectiveness of a method, and a number of studies have been conducted to establish the clinical utility of PAC.
Connors et al. in a multicentre RCT demonstrated a relative increase in mortality of 26% and an increase in hospital costs by 38% associated with PAC use in 5,735 critically ill patients [
A commonly cited benefit of PAC is that it provides filling pressures which can be used to identify fluid responsiveness and guide fluid administration. However, these filing pressures have been found to be neither uniformly accurate [
The accuracy of CO/SV measures determined by USCOM is dependent on the accuracy of the prediction of the valve area and any error in this calculation will convert directly to an error in CO/SV. Two studies have reported poor agreement of transthoracic and transesophageal echocardiographic measurement of the aortic valve (AV) and pulmonary valve (PV) diameter compared with the morphometrically determined USCOM values. These studies consequently reported poor agreement of echocardiographic, and PAC determined CO values [
The outcomes benefits of hemodynamic optimization seem intuitive and are the rationale for circulatory interventions with fluid, inotropes, and vaso-active therapies. In presurgical patients hemodynamic optimization has been demonstrated in a 29 study meta-analysis to produce a decrease in overall mortality (7.6%) and morbidity [
The PAC is used for CO measurement in sheep and as a cardiovascular research tool, while the USCOM algorithms were developed in human subjects. To compensate the USCOM device was calibrated to baseline FP measures, and this may have conferred some benefit to the FP/USCOM bias comparison. However, calibration data were excluded from analysis and 76% of all measurements were made after interventions which altered CO across a sixfold range (1.9 to 11.7 L/min), thus mitigating any conferred benefits from calibration. Additionally the interventional component of the study, designed to assess sensitivity to CO change, involved calibrating all methods to a nominal zero baseline, thus removing any advantage for any method. As a further precaution a calibrated PAC series, cPAC, was generated post hoc to address any methodological bias in the study.
We found a 20% failure rate of PAC during the monitoring of normal sheep. We hypothesised that as CCO PAC has a thermometrically triggered safety cutoff, and that the normal body temperature of sheep is 39° to 39.9°C [
This study was of a continuous CO iteration of PAC, CCO, and so some of the observations may not be interchangeable with bolus thermodilution. However, CCO is a thermodilution method and has been adopted on the basis of very good to excellent agreement with iced bolus thermodilution [
The FP was implanted around the ascending aorta measuring left heart CO while both PAC and USCOM measured right CO; however, in the absence of a shunt and disregarding coronary flow, the values should equate and the comparison remain valid.
The PAC was sited in the pulmonary artery throughout the studies, and would have caused some disturbance to the flow characteristics measured by USCOM. This limitation of the catheter measurements cannot be overcome in the current preparation, and would also be a confounding factor in clinical practice.
This study found that PAC was neither accurate nor sensitive when compared with the FP, findings which may explain the apparent ineffectiveness of PAC in clinical practice. This study also found that USCOM provided equivalent CO measurements to FP and is a noninvasive and accurate alternative to PAC reliably detecting ±5% changes in CO.
An USCOM 1A device was lent to the Howard Florey Institute for the purposes of this study. R. A. Phillips and B. M. Jacobson are shareholders and employees of Uscom Ltd.
Thanks are due to Simon Parker for statistical support, and Brendan Smith, Joe Brierley, David Bennett, Darryl Burstow, and John Fraser.