1 Introduction

Cardiac output (CO) measurement constitutes an integral part of the diagnostic and therapeutic strategy in critically ill patients. Measuring CO allows physicians to characterize shock states in patients and to guide therapy at the bedside, especially during fluid resuscitation and adaptation of vasoactive or inotropic support.

During the last decade, transpulmonary thermodilution (TPTD) using the single indicator technique [1] was implemented in clinical practice and gradually replaced the pulmonary thermodilution. The development and the implementation of this new technique (TPTD) in clinical practice are significantly reflected by the number of publications in Index Medicus between 1982 and 2014. A total of 349 citations appear when the keyword “transpulmonary thermodilution” is typed.

The TPTD curve allows the measurement of CO (COTPTD) and the estimate of derived indices, thereby permitting assessment of cardiac preload [2, 3] and extravascular lung water [4–6]. COTPTD is determined using the Stewart-Hamilton equation applied to a thermodilution curve [7], which is similar when using a pulmonary artery catheter [8]. Compared to the right heart thermodilution curve, a notable exception lies in the injection site of the thermal indicator and the recording site of the thermodilution curve: the TPTD curve is obtained by injecting the thermal indicator into a central intrathoracic vein instead of the right atrium, and the thermal shift is collected via an arterial catheter inserted in a large systemic arterial trunk (femoral, axillary or brachial artery), instead of the pulmonary artery, after a transpulmonary transit of the cold indicator.

At its output, the TPTD technique was the subject of debate concerning whether this new technique could be a surrogate of CO measurement by pulmonary artery catheter. In this regard, the evaluation of the reproducibility of the present technique is essential for recognizing the difference between changes in the patient’s physiological status and errors associated with the technique. However, only one study [9] has specifically evaluated the reproducibility of COTPTD and the assessment of how closely individual measurements agree with each other when all efforts attempt to keep hemodynamic conditions constant.

The purpose of this paper is to systematically review and critically assess the existing data concerning the reproducibility of COTPTD. The present analysis may, in large part, impact clinical practice and the definition of a positive response to fluid infusion. The ‘magic threshold value’ of a 15 % increase in CO, which is adapted to assess fluid responsiveness, is primarily related to the precision of the pulmonary artery catheter [10].

2 Methods

The statistical definition of precision is the variability of values attributed to random measurement errors. Precision means repeatability when the technique of measurements is performed by the same operator, during constant conditions, and repeated within a short time period. Precision equals reproducibility when the process of measurements is performed among different operators (laboratories) and over longer time periods. Thus, when performing a systematic analysis and adding repeatability of COTPTD values obtained from distinctive studies, the computation equals the measurement of reproducibility.

2.1 Search strategy

Combinations of key words related to “Transpulmonary thermodilution AND precision”; “Transpulmonary thermodilution AND reproducibility”; “Transpulmonary thermodilution AND CO” were used to search reviews and reports on the MEDLINE and Embase databases and the Cochrane Group trial register without any language restriction. The number of articles published in each category was 25, 52, and 234, respectively. The last search was performed in December 2014. We checked the bibliographies of the retrieved reports and reviews. For conducting and reporting the present systematic review, the authors followed the MOOSE Guidelines for observational studies [11], when indicated. As there is no gold-standard validated method for assessing bias and the quality of observational studies (10, [12] but there is consensus that the three most fundamental domains are (1) appropriate selection of studies, (2) appropriate measurement of variables, and (3) appropriate control of confounding, as well as considering design-specific biases, we therefore discuss these quality criteria for each study in a narrative fashion. Also, we did not consider data from abstracts, letters and animal studies. Because this study did not involve human participants or medical records, institutional review board approval was neither required nor obtained.

2.1.1 Inclusion criteria

To assess the reproducibility of TPTD technique in all of the studies, we assessed the precision reported in each study (repeatability). The precision was often referred to the coefficient of error of the estimate, which is a statistical expression for the size of the standard error of the mean of repeated estimates, relative to the mean of the estimates. In cases where we were unable to find the COTPTD precision value, we extracted either the coefficient of variation (CV), or the coefficient of error (CE), or the standard deviation of the mean (SEM), or the least significance change (LSC). CE and CV were obtained using the formula: CE = CV/√n or CV = CE × √n, respectively, where n is the number of replicates of measurements in each patient. The precision is calculated as two CEs. A measurement precision level of ≤10 % is desirable [13]. The LSC is the minimum change that must be measured by a device to recognize a real change of measurement [14]. LSC was calculated using the following equation: LSC = CE × 1.96 × √2 [9].

After reading and analyzing these articles, 16 studies (n = 14 adult studies and 2 pediatric studies) that met the inclusion criteria were identified (Fig. 1). An additional computerized search, “ISI Web of Knowledge”, which analyzed citing and cited articles as well as a manual search in the references of relevant articles, did not uncover any more articles that could be included in our systematic analysis. The characteristics of all of the studies were independently rated by R.G. And K.B., and the discrepancies were discussed.

Fig. 1
figure 1

Flow diagram describing the data analysis protocol

Full size image

We found 14 studies in which TPTD was used to measure CO, without including reproducibility results in their articles. We attempted to contact these eight research teams to obtain additional data. Unfortunately, only one response was sent to us with unusable data in the context of our systematic review. In selected studies for this systematic review, we wanted to obtain additional data, such as the standard deviation (SD) of CO measurements, to perform a meta-analysis with complete results. We contacted the authors of these studies (14 e-mails were sent twice a month apart); however, we received only two responses.

In each case, the number of measurements for determining a CO value, number of CO averaged measurements, temperature, nature, volume and injection site of the indicator and insertion site of the arterial catheter were collected. Additionally, the number of cold boluses to obtain a CO averaged value and the total numbers of thermodilution curves were required for calculating reproducibility.

2.2 Statistical analysis

The results are expressed as absolute values for the number of averaged CO measurements and for the total thermodilution curves achieved, as the mean ± SD and the median for CO measurements or cardiac index and as percentages for precision, CV and LSC. Graphs were constructed using GraphPad Prism 5.0 (GraphPad Software, San Diego, CA, USA) and Microsoft office Excel 2007.

3 Results

A flow diagram is shown in Fig. 1, and demographics data are presented in Table 1 for adult studies and in Table 2 for pediatric studies. In these tables, we wanted to show CV, LCS and precision and to clarify which values were extracted from specific articles to achieve the present results.

Table 1 Articles from adult studies used for data analyses, patient characteristics, TPTD parameters, and results of the data analysis concerning the coefficient of variation, coefficient of error, precision, least significant change of the TPTD technique, number of boluses, number of average cardiac output measurements and total number of boluses in each study
Full size table
Table 2 Articles from pediatric studies used for data analyses, patient characteristics, TPTD parameters, and results of the data analysis concerning the coefficient of variation, coefficient of error, precision, least signification change of the TPTD technique, number of boluses, number of average cardiac output measurements and total number of boluses in each study
Full size table

The primary results are shown in Tables 1 and 2. A total of 3432 averaged CO values (9775 TPTD boluses) were reviewed from 14 adult studies and 78 averaged CO values (210 TPTD boluses) were reviewed from 2 pediatric studies. In all studies CVC were positioned in the superior vena cava. The average number of bolus made by TPTD to obtain a CO value was 3.3 ± 0.6 in the adult studies and 2.7 ± 0.6 in the pediatric studies. In 12 of 14 adult studies and in 2 pediatric studies, the authors conducted three thermodilution boluses; in 2 adult studies, they performed 4 boluses; in 1 adult study, they performed 5 boluses; and in 1 adult study, they performed 2 boluses.

Concerning the study conducted by Kiefer et al., the same 72 patients received 3 boluses to assess a CO value. The TPTD curves were analyzed twice: on the one hand, on-line using the EV1000TM algorithm; and on the other hand, off-line, using the PiCCO TM2 algorithm (this study appears twice and is counted twice, as shown in Table 1).

The overall reproducibility of COTPTD was 6.1 ± 2.0 % and LSC was 8.5 ± 2.7 % for CO measurements from the 14 selected adult studies. The overall reproducibility of COTPTD was 3.9 ± 2.9 % and LSC was 5.4 ± 4.0 % for CO measurements from the two selected pediatric studies (Table 2) (Fig. 2a, b).

Fig. 2
figure 2

Aligned plots representing the precision (a) and LSC (b) of adult and pediatric studies

Full size image

In the present study, we performed a subgroup analysis after classifying the data into three categories based on the number of boluses injected to determine an average CO value: less than 3 boluses, 3 boluses and more than 3 boluses. Reproducibility was 6.6 ± 1.7, 5.4 ± 1.7 and 10.2 ± 3.7 %, depending on whether 3, more than 3 or less than 3 boluses were performed, respectively. LSC was 9.1 ± 2.3, 7.5 ± 2.3 and 14.2 ± 5.1 %, depending on whether 3, more than 3 or less than 3 boluses were performed, respectively (Fig. 3a, b).

Fig. 3
figure 3

Aligned plots representing the precision (a) and LSC (b) whether less than 3 boluses, 3 boluses and more than 3 boluses were administered to realize the transpulmonary thermodilution

Full size image

The cold boluses were injected through the proximal port of the pulmonary artery catheter (PAC) in 5 studies or through the central venous catheter (CVC) in 9 studies. In one study, it was not apparent in which site the cold boluses were injected. The reproducibility and the LSC of the COTPTD measurements were not affected by the injection site (Supplemental Digital Content 1 a and b). In 12 studies, the PiCCO arterial catheter was inserted in the femoral artery, whereas in 2 studies, the arterial catheter was inserted in the axillary artery (Supplemental Digital Content 2 a and b).

The amount of cold injected liquid was variable according to the studies. The injected liquid ranged from 10 to 20 mL, knowing that in 10 studies, the cold bolus was at least 15 mL. The injection site did not seem to influence the reproducibility nor the LSC of the CO measurement by TPTD (Supplemental Digital Content 1. a and b and 2. a and b). However, when the amount of fluid injected and the site of injection were analyzed together (CVC 15 mL vs. CVC 20 mL and PAC 10 mL vs. PAC 15 mL), the precision and the LSC were better for both the injection of 20 versus 15 mL cold liquid through the CVC and the injection of 15 vs. 10 mL via the PAC (Fig. 4a, b). Moreover, in all of the studies, the cold bolus was a liquid with a temperature below 8 °C.

Fig. 4
figure 4

Aligned plots representing the precision (a) and LSC (b) whether 15 or 20 mL was injected through the central venous catheter (CVC) or 10 or 15 mL was injected through the proximal port of the pulmonary artery catheter (PAC) to realize the transpulmonary thermodilution

Full size image

Finally, we constructed a bubble graph, which illustrates that the precision of TPTD depends on the number of boluses to calculate a CO value and the number of boluses achieved in each study. The bubble graph showed that reproducibility improves with an increased number of boluses (Fig. 5).

Fig. 5
figure 5

Bubble graph illustrating the precision of TPTD assessed in each study depending on the number of boluses achieved to calculate a CO value. The surface of each bubble represents the magnitude of boluses made in each study and considered in this review. Pink bubbles adult Studies, Blue bubbles pediatric studies

Full size image

4 Discussion

The primary finding of the present systematic review is that the TPTD technique has a higher reproducibility value compared to the pulmonary artery catheter [1, 10]. This result is likely related to the longer transit time of the thermal bolus, which finally is not influenced by respiration and arrhythmias.

To detect 15 % change in CO, a monitor must have a level of precision that can detect this change with 95 % certainty. However, measuring a change does not necessarily mean that the patient’s physiological status has changed [15]. The reproducibility of a technique is its ability to provide CO values that are close to each other. This factor is especially important for techniques measuring CO [16] because we are more interested in variations of CO values over time than in a given CO value. In this regard, reproducibility of a technique capable of measuring CO is important, particularly when the task is to assess fluid responsiveness. For example, when a patient is considered to be fluid responsive because his CO increases above 15 %, it is essential to ensure that the difference in CO measured before and after volume expansion is actually the patient’s physiological reaction to the volume administration, rather than related to the error associated to the measurement technique. As the magic threshold value of 15 % CO increase is primarily related to the clinical significance and reproducibility of the pulmonary artery catheter [10, 17], we could argue that when CO is measured by TPTD, a technique that has a reproducibility of 6.1 ± 2.0 %, a 10 % CO increase could define a patient to be fluid responsive.

The good reproducibility of TPTD can be explained by the fact that using this technique, the cold indicator takes approximately twenty-seconds to be captured by the femoral temperature sensor, whereas using the right heart thermodilution, the temperature variation is detected after approximately 3–4 s. The present longer transit time reduces artefacts produced by airway pressure and arrhythmia on venous return-CO steady state, which have more impact on the results of right heart thermodilution [18, 19]. Also, the relative very good reproducibility observed in the pediatric population [20, 21] can be explained by several mechanisms: first, the small size of children allows arterial and venous catheters to be inserted specifically in the thoracic compartment. Second, the heart rate and respiratory rate are higher for children than for adults, thus inducing a lower influence on the TPTD curve. Finally, these results support the use of TPTD for measuring CO in pediatric patients because this technique can be easily performed in children, unlike the pulmonary artery catheter [22].

A mean of 3 measurements is recommended by the manufacturer for determining CO [23]. However, in practice, some teams realize 2 or even only one measurement for determining CO. Our results suggest that injecting more than 2 boluses may improve the reproducibility of TPTD measurement. However, when the amount of fluid injected and the site of injection were analyzed together, the reproducibility and LSC were improved when a higher volume was injected. The injection of 15 mL of iced-cold saline is recommended by the manufacturer for each TPTD bolus. Our results showed that a higher volume of cold indicator injected provides a better precision with an improvement in the reproducibility. Indeed, a potential loss of the thermal indicator is a well-known weakness of any thermodilution technique due to thermic exchange between the indicator and the surrounding anatomic structures (vessels, valves, heart chambers, etc.). [24]. Compared with pulmonary arterial thermodilution, the thermodilution bolus in TPTD is more prone to indicator loss with regard to the longer transit time of the thermal bolus and to an increased contact surfaces in the pulmonary circulation. Increasing the amount of cold indicator volume (20 mL) are likely to better compensate for indicator loss.

The present review has a number of limitations. First, the number of studies from which we could extract the data on TPTD reproducibility remained low compared with all of the studies published using this technique. Second, patients included in different studies are not homogeneous in terms of medical and technical conditions. Third, this systematic review included studies published over the last 12 years. During this period, the PiCCO™ software has been edited twice: PiCCO™, PiCCOPlus™ and PiCCO2™ could impact the reliability of our results. Fourth, we had not included the lithium dilution technique (LiDCO Ltd, Cambridge, UK) in the present systematic review as the aim was to measure the precision of transpulmonary thermodilution method and not the precision of transpulmonary lithium-dilution technique. Fifth, the use of TPTD in pediatric patients could be tempered. Only two small studies (30 patients in all) have been included in this systematic review. Finally because our study is a systematic review without meta-analysis, we only conducted a graphical representation of results.

5 Conclusion

Our results showed that TPTD has a very good reproducibility compared to the right heart thermodilution. Moreover, the present study may suggest that the definition of a positive response to fluid infusion fixed at 15 % due to the lack of reproducibility of PAC could be redefined at only 10 % when TPTD is used. We believe that this 10 % change in CO could be statistically and clinically relevant.