This prospective analysis of 927 hours of data from post-operative cardiac surgery patients recovering in the intensive care unit showed a moderate correlation between the gold standard of measurement (CO-CTD via the pulmonary artery catheter) and the CO-MBA method using the peripheral arterial waveform.
Intermittent pulmonary artery thermodilution uses injection of several fluid boluses with a known volume and temperature into the right atrium. The temperature change detected by a distal thermistor is then used to calculate the cardiac output [19]. In contrast, continuous pulmonary artery thermodilution uses a thermal filament in the right ventricle, without the need for manual injections [19]. This method has been proven to be as accurate and reliable as standard bolus thermodilution, and both methods are used clinically to manage the hemodynamic status of critically ill patients [20]. This report is the first analysis of agreement between the continuous cardiac output Swan-Ganz catheter (CO-CTD) and the novel multi-beat analysis of arterial waveform (CO-MBA). Animal data and retrospective analysis from ICU data comparing cardiac output with the CO-MBA against a reference calibrated standard have reported good agreement.[7, 21, 22]
A debate about the merits of PAC placement has been ongoing with multiple large scale studies both showing and failing to show an overall clinical benefit to placement [1]. Adverse events ranging in severity from self-limiting arrhythmia to pulmonary artery rupture have been reported in the literature with an incidence from two to seventeen percent [2]. Less invasive methods of cardiac output determination, such as the CO-MBA, may allow the intensivist to avoid the PAC and supplement it in other situations where extended periods of monitoring may be necessary. Multiple studies have compared various pulse contour cardiac output methods to intermittent bolus thermodilution cardiac output via the PAC and have shown good correlation, concordance, and lower percentage errors [23–28] Accuracy of measurement is critical, and our analysis showed a percentage error of 38.2% and a mean of differences 0.04 ± 1.04 L/min, 95% limits of agreement: -2.00 to 2.08 L/min, in the full cohort and a percentage error of 35.2% and a mean of differences 0.14 ± 0.90 L/min, 95% limits of agreement: -1.63 to 1.91 L/min in the in the arrhythmia cohort respectively. Although the CO-MBA agreement with CO-CTD reported here, is outside the 30% error limit set by the well-known Critchley and Critchley analysis [18] other meta analyses of a combination of calibrated and uncalibrated methods have found that the error in clinically used pulse contour devices is about 42%. [29]
Greiwe and colleagues also specifically compared the CO-MBA during off pump coronary artery bypass graft surgery and in the intensive care unit using intermittent thermodilution in a method comparison study afterwards.[9] They had a smaller sample size than our report ( 167 comparison points; 31 patients ) and performed bolus thermodilution cardiac output measurements at seven pre-defined time points. Furthermore, this was not a real time data collection with a bedside monitor, rather was an offline analysis where arterial blood pressure waveforms were fed into the Argos monitor retrospectively. Percentage error reported was 40.7% and not very different from the 35–38% range in our cohorts. Our results are also consistent with the CO-MBA technique method comparison analysis reported in the cardiac operating room by Saugel and colleagues [10]. CO estimations showed reasonable agreement and trending ability between the two methods, with a concordance rate of 89%. [10] Another recent study comparing the Argos and FloTrac monitors, showed that Argos was more accurate with a higher concordance rate, and thus may prove valuable in CO trending. Although, both devices were not interchangeable with thermodilution for absolute CO measurement due to high percentage errors of 50% .[30] We identified 10 subjects (11%) as having consistently underdamped blood pressure signals on the arterial line and were excluded. For comparison, other investigators have previously excluded 9 out of 40 subjects (23%) and found 92 out of 300 subjects (31%) to have underdamping or resonant arterial lines in similar patient populations. [9] [14]
Up to a third of patients have atrial and ventricular arrhythmias after cardiac surgery.[31] Some of these rhythm changes last for a significant amount of time and are associated with hemodynamic instability. This may result in periods of under perfusion and trigger organ system injury We specifically analyzed a subgroup of patients with rhythm disturbances identified by two anesthesiologists and found moderate correlation here as well. The reliability of cardiac output measurements during rhythm instability shown in this work is of critical significance. Even though we only had 26 patients in this subgroup, our work serves as hypothesis generating for future analysis where specific types, times and durations of rhythm changes could be analyzed in larger cohorts.
This work is novel and has several advantages. We had a large dataset with highly granular comparative data with CO-MBA measurements recorded once every 5 seconds, and CO-CTD recorded once every minute, generating 55599 CO-CTD and CO-MBA data pairs, and over 900 hours of usable data. Furthermore, we performed post processing, where we averaged out the CO-CTD over an hour and the CO-MBA to once every minute, to compensate for the difference in response time between the CO-MBA and the continuous cardiac output swan-Ganz. Patients were prospectively enrolled into the study, with broad inclusion criteria, and data for the CO-MBA and the CO-CTD was recorded real-time at the bedside. The CO-MBA was blinded, alarms silenced, and clinicians followed standard management protocols with the pulmonary artery catheter. Our bedside ICU nurses routinely check the integrity of the arterial line using the ‘fast-flush test’ which allowed us an assurance that the data generated was accurate.[13, 32, 33] In addition, visual inspection of available data by two experienced anesthesiologists let us exclude segments of over damped or under damped arterial line morphology. Rhythm disturbances being common after cardiac surgery, this analysis was the first comparison of agreement with a subgroup of patients with these elements, in addition to being the first analysis to use the continuous cardiac output from the PAC as a comparator.
There are some limitations to this analysis as well. While we allowed for adjustments for CO response time differences between the two comparison methods by averaging both methods over a one-hour time scale, this may have the effect of reducing the variance in both methods. We are unsure of the clinical validity and relevance of an earlier detection of a change in cardiac output, however, our delay in response time reported ( 12–24 minutes ) for the continuous thermodilution swan-Ganz is not different from published work reporting a response time of 10 minutes or more. [15]Furthermore, the arrythmia subgroup involved a small sample size, was of uncertain clinical significance including the actual hemodynamic changes involved, and the specific nature or duration of the rhythm disturbances. This was novel all the same, and relevant since rhythm disturbances are very common in the post-operative cardiac surgery patient. While removing artifactual and underdamped or overdamped arterial line waveforms and data was crucial, we could have introduced a selection bias and made this data non-representative of the real world where such periods of less-than-optimal monitoring constitute a non-trivial part of the total monitoring time of a critically ill patient in the ICU. In CO method comparison studies with intermittent thermodilution as the reference method, a four-quadrant concordance analysis is often suggested as a way to assess the trending agreement. Here, because the reference method was continuous thermodilution with consecutive samples being one minute apart, a meaningful four-quadrant analysis could not be performed. In addition, we did not plan for comparisons at specific interventions such fluid loading or rapid changes in inotropy or afterload and could not perform a trending analysis. Finally, this work comes from a single center and may be reflective of local practice patterns precluding generalizability.