Introduction

Haemodynamic instability is a common problem in ICUs worldwide, both in patients admitted as emergencies and after elective major surgery. Handling this instability appropriately involves realising that it is present in the patient, diagnosing its causes and choosing the correct treatment options. ^{Reference Bentzer, Griesdale, Boyd, MacLean, Sirounis and Ayas1} Unfortunately, clinical examination and traditional laboratory analyses have been shown to struggle to identify patients with deranged haemodynamics, and picking the right remedy based on clinical parameters has proven to be difficult, even for experienced clinicians. ^{Reference Hiemstra, Koster and Wiersema2} A search for a more comprehensive and objective method for evaluating the haemodynamics of critically ill patients has thus been going on for many decades. Initially, the pulmonary artery catheter was used. ^{Reference Arias-Ortiz and Vincent3} While it provides much useful data that other methods cannot give, its use has diminished since the advent of less invasive methods. There are also conflicting data regarding its usefulness outside of cardiothoracic surgery and intensive care, and a lively debate over its place in current clinical practice has been ongoing for a while. ^{Reference Szabo, Betances-Fernandez, Navas-Blanco and Modak4,Reference Navas-Blanco, Vaidyanathan, Blanco and Modak R.K.5}

Measuring cardiac output is one of the most useful variables that a pulmonary artery catheter provides, and the search for a method which can deliver a similar measurement without necessitating a catheter in the pulmonary circulation has provided the ICU with the transpulmonary thermodilution method. ^{Reference Monnet and Teboul6} Two manufacturers provide similar apparatuses, and in this study, we have used the PiCCO™ equipment.

Transpulmonary thermodilution is thus in many ways similar to the pulmonary artery catheter. As a replacement or complement, echocardiography is often used. This method may be used to estimate the cardiac output, but can also give detailed information on cardiac contractility, valvular lesions, and cardiac volumes, such as the end-diastolic volumes of the ventricles. ^{Reference Tissot, Singh and Sekarski8} A major drawback of echocardiography, though, is that it only provides a snapshot of the circulation at one time; another is that its accuracy depends on the personal skill of the clinician. Thus, the possibility of having access to a continuous cardiac output measurement, which is not dependent on which staff is available, is an attractive idea. The question, of course, is how accurate the transpulmonary thermodilution estimate of the cardiac output really is.

Cardiac surgery for CHDs is a procedure which often causes haemodynamic instability during the first 24 postoperative hours, particularly in neonates and young infants. ^{Reference Bronicki and Chang9,Reference Aslan and Yildizdas10} These patients are much too small in size to be monitored using a pulmonary artery catheter, which means that other methods must be considered. We decided to evaluate the use of a transpulmonary thermodilution setup with echocardiography examinations as a comparison. A relatively homogenous patient cohort was chosen, namely children under the age of one year, scheduled for elective repair of a ventricular septal defect, or an atrio-ventricular septal defect, all of which had a non-restrictive blood flow over their septal defects which caused mild to moderate heart failure preoperatively. The aim was thus to compare cardiac output values using simultaneous measurements with a transpulmonary thermodilution method and echocardiography, with the null hypothesis that the methods would yield comparable results.

Materials and methods

Results

Study population

All children completed the study, and the data of all the 40 patients were included. There was no in-study mortality. The median age was 160 (range 133) days, and median weight was 5.6 (range 4.3) kg. There were 14 female and 26 male patients, and 24 had Down’s syndrome (trisomy 21). Mean preoperative Ross score was 6.1 (standard deviation 2.0). 23 patients had a Ross score between 3 and 6, indicating mild heart failure, and 17 had a score between 6 and 9, indicating moderate heart failure. For a summary of the patient characteristics, see Table 1.

Table 1. Patient characteristics

SD = standard deviation, AVSD = atrioventricular septal defect, VSD = ventricular septal defect, CPB = cardiopulmonary bypass.

Cardiac output at 18 hours after surgery

There were small differences between the median of the two methods. The median cardiac output for PiCCO™ was 1.16 l/min (range 0.49–1.71 l/min) and the median cardiac index 3.63 (range 2.13 – 5.91) at 18 h after surgery. The median cardiac output for echocardiography was 1.05 l/min (range 0.42–1.66 l/min) and the median cardiac index 3.30 (range 1.20–5.90). A box plot of the cardiac output can be seen in Figure 1. There was no statistically significant difference between the methods neither in cardiac output nor in cardiac index.

Figure 1. CO = cardiac output.

Agreement of COcardiac output measurement between PiCCO™ and echocardiography

The agreement of the two methods for measuring cardiac output was analysed 18 h after the end of surgery. The samples were not normally distributed, and calculations were made using log-transformed data. PiCCO™ yielded on average 3.0% (95% CI -10.3–24.1%) higher values than echocardiography. 95% of the observations fell within -50.0 and 82.6% difference. The results are presented as a Bland–Altman plot in Figure 2.

Figure 2. Dotted line: mean difference (%). Dashed lines: interval of 95% of the observations. Shaded areas: 95% CI for mean difference and the upper and lower 95% intervals. CO = cardiac output.

Discussion

In this sample, the analysis showed a small systematic bias when measuring cardiac output with transpulmonary thermodilution and echocardiography, evident by the modest mean difference between the methods. However, the spread of the results was large, with individual sample values exhibiting considerable differences between the two methods. The magnitude of the variation across the range of mean cardiac output in the study, visible in the Bland–Altman plot, shows no tendency to change with increasing or decreasing cardiac output.

There have been previous attempts to evaluate how well different methods of haemodynamic monitoring perform in clinical settings, with conflicting results. Most studies of paediatric haemodynamics focus on the responses to fluid boluses, trying to find methods to identify responders to fluid therapy in haemodynamically unstable patients. ^{Reference Shostak, Shochat, Manor, Nahum, Dagan and Schiller12,Reference Yenjabog, Kanchongkittiphon, Chutipongtanate, Lertbunrian and Ungprasert13} This study is not designed to evaluate any haemodynamic interventions, however. In this study, the physicians responsible for the care of the patient did not have access to the information gathered by the two methods, to minimise the risk of bias in the treatment of the patients.

Two previous studies relate relatively closely to ours. Gergely et al ^{Reference Gergely, Ablonczy and Kramer14} studied infants after cardiac surgery of the same age as the present study did, albeit with different and varying diagnoses. They could not find any correlation between cardiac output measurement using transpulmonary thermodilution and echocardiography. This study was quite small, however, and no estimate of agreement using a Bland–Altman plot was provided, something that is recommended for method comparisons. ^{Reference Martin Bland and Altman15} Aslan et al ^{Reference Aslan, Yildizdas and Horoz16} compared PiCCO™ and echocardiography in another small study of children with non-cardiac haemodynamic instability (such as septic shock and burns). In this study, the correlation between the methods was excellent, and a Bland–Altman plot revealed a small mean difference, with a relatively high variability of individual measurements. These results are similar to those in our study; however, the variability between the individual patients in our data is even greater. It is possible that this is related to patient age (the patients in the study by Aslan were much older) or clinical situation (our patients being monitored after cardiac surgery).

There are inherent advantages and drawbacks with both transpulmonary thermodilution and echocardiography for haemodynamic monitoring. Echocardiography is non-invasive, and there are no risks for the patient during the procedure. The method only provides a moment frozen in time of the patient´s status, however, which can change rapidly. A competent clinician for performing the examination is also required, something which may not be available at all times (although intensivists who are not cardiologists can be trained in focused echocardiography for the limited purpose of haemodynamic evaluation ^{Reference Tissot, Singh and Sekarski8} ). In a postoperative setting, limited access to the patient because of bandages, cables, and on occasion an unclosed sternum may also compromise an ideal examination.

Transpulmonary thermodilution methods (in this case the PiCCO™ system) also mandate a skilled clinician for the insertion of the catheters; however, after this, the use of the method necessitates no particular technical competence. The system also provides the possibility of frequent measurements, with pulse contour analysis available between calibrations for a continuous assessment of the cardiac output. The precision of the transpulmonary thermodilution has been found to be decent, with the ability to discern changes in cardiac output down to 12%. ^{Reference Monnet, Persichini, Ktari, Jozwiak, Richard and Teboul17} Transpulmonary thermodilution methods are also able to stand up quite well to the gold standard of the pulmonary artery catheter, with cardiac output measurements using the different methods being equally useful. ^{Reference Stetz, Miller, Kelly and Raffin18} There are pitfalls in this method too, though. In a setting of paediatric cardiac surgery, the most obvious concern is that most infants and children who are scheduled for cardiac surgery have intracardiac shunts of varying degrees, something which makes the transpulmonary thermodilution measurements highly unreliable because of recirculation of the cold saline. In the present study, no transpulmonary thermodilution measurements were made before the surgery for this very reason; most of the patients had a pulmonary blood flow of at least 2–3 times that of the systemic circulation. Also, in contrast to echocardiography, transpulmonary thermodilution provides no direct assessment of the cardiac contractility, which is often important information in the first 12–24 h after cardiac surgery using cardiopulmonary bypass. There is also no way of measuring the filling pressures of the right and left ventricle separately, which is possible with a pulmonary artery catheter.

A drawback with the PiCCO™ system in infants which we have observed is the risk of inadequate arterial circulation in the leg into which the catheter has been inserted into the femoral artery. The paediatric catheters provided by the manufacturer are smaller than the adult versions, but still occlude a relatively larger cross-sectional area of the blood vessel than in adults. No serious complications arose in our study, but the risk of ischaemia is still one of the reasons that the PiCCO™ system is not used more frequently in our centre.

The question is, of course: which of these methods, if any, is appropriate to use after paediatric cardiac surgery for estimating cardiac output? On an aggregate level, the methods display excellent agreement, as can be seen by the small mean difference between them. However, the large variation between the methods in individual patients does present a problem for the clinician. It is clear that if trends are to be followed in a patient, the methods are not interchangeable, and cannot be compared directly.

Conclusions

The agreement of cardiac output measurement between echocardiography and the transpulmonary thermodilution method (here, PiCCO™) was strong. However, the variation of the difference between the two methods in individual patients was considerable. The results from this study suggest that even though both methods may be useful for monitoring infants after cardiac surgery, they cannot be used interchangeably for cardiac output measurement in a clinical context.