Experimental Validation of Cardiac Index Measurement Using Transpulmonary Thermodilution Technique in Neonatal Total Liquid Ventilation : ASAIO Journal

Secondary Logo

Journal Logo

Pediatric Circulatory Support and Perfusion

Experimental Validation of Cardiac Index Measurement Using Transpulmonary Thermodilution Technique in Neonatal Total Liquid Ventilation

Bossé, Dominick*; Walti, Hervé*†; Robert, Raymond; Lebon, Johan*; Lesur, Olivier; Praud, Jean-Paul*†; Micheau, Philippe

Author Information
ASAIO Journal 56(6):p 557-562, November 2010. | DOI: 10.1097/MAT.0b013e3181f1cd72
  • Free


Pulmonary artery thermodilution (PTD) is one of the best currently available methods for neonatal cardiac index (CI) monitoring that is both bedside feasible and sufficiently rapid to be used by clinician.1 Nonetheless, reluctance of intensivists to introduce a catheter into the pulmonary artery,2 with its share of complications,1,3–5 has led to the development and validation6 of the less-invasive transpulmonary thermodilution (TPTD) method, which uses femoral instead of pulmonary artery measurement. Fortunately, even if TPTD requires that thermal indicator passes through the pulmonary circulation, it has been proven that simultaneous use of mechanical ventilation does not alter CI measurement in children and infants.1 However, the development of liquid-assisted ventilation for the management and treatment of neonatal pulmonary disease such as infant respiratory distress and meconium aspiration has prompted the re-emergence of this question.

Two modes of liquid-assisted ventilation7 have been developed: 1) total liquid ventilation (TLV), where the lungs are completely filled with a perfluorocarbon liquid (PFC) while a dedicated liquid ventilator ensures the renewal of a tidal volume of liquid and 2) partial liquid ventilation (PLV), a simplification of TLV, where the volume of PFC instilled into the trachea is close to or less than the functional residual capacity of gas-ventilated lungs. PLV mainly differs from TLV by using a conventional gas ventilator and by maintaining an air-liquid interface. The considerable volume of heated PFC present in the lungs during liquid-assisted ventilations and the much greater heat capacity of PFC (1680 times higher that air) has led to hypothesize a theoretical loss of thermal indicator within liquid-filled lungs that could lead to a systematic overestimation of CI by TPTD. Conversely, pulmonary sensor is barely or not influenced by mediastinal temperature and PLV would have minimal effects on PTD reliability.8,9

Validation of CI monitoring devices such as TPTD and PTD is an integral part of the development of TLV, and to the best of our knowledge, the precision of both thermodilution techniques in TLV has yet to be addressed. Therefore, as part of a larger study of the efficacy and cardiovascular consequences of TLV in comparison with conventional mechanical ventilation (CMV) in an animal model of acute respiratory distress syndrome,10 we ascertained that precision (repeatability) of both thermodilution measurements was preserved under TLV and sought to determine whether interchangeability was maintained between TPTD and PTD.

Materials and Methods

The experimental protocol was approved by our institutional Ethics Committee for Animal Care and Experimentation.

Animal Preparation

Seventeen term newborn lambs, weighing 4.2 ± 0.4 kg and <5 days of age, were first premedicated, ventilated, and instrumented as previously described.11 Anesthesia, analgesia, and curarization were achieved according to the same protocol. Antibiotics (0.05 mg/kg duplocillin and 5 mg/kg gentamycin) were injected intramuscularly, and a continuous infusion of 4 ml/kg/h of 5% dextrose was administered via the jugular vein throughout the experiments. Pulsion Medical System's (Munich, Germany) 6Fr (PV2047, VoLEF) and 3Fr 7-cm (PV2013L07, PiCCO) thermodilution catheters were positioned into the pulmonary artery and right femoral artery to perform PTD and TPTD, respectively.

Study Design

Lung Injury Protocol and Randomization.

Before lung injury, FiO2 was increased at 1.0, and preinjury data were collected after a 30-minute stabilization period. Thereafter, lung injury was established by instillation of 10 ml/kg of warm hydrochloric acid in saline (37°C; pH 1.2–1.6). The solution was instilled into the lungs via the endotracheal tube by gravity, with a gradient of 20 cm H2O. After instillation, the lamb was reconnected to the ventilator and positive end-expiratory pressure (PEEP) was increased to 9 cm H2O for 1 minute. The lamb was then disconnected from the ventilator and the acid solution was drained from the lungs by gravity and gentle suction. Serial lung lavages were performed at 15–20 minutes intervals until at least a 50% decrease in PaO2 from baseline measurements was obtained. After lung injury, postinjury data were collected and lambs were randomized in to three groups: conventional gas ventilation (n = 6); TLV with perfluorooctylbromide (PFOB) (n = 5); and TLV with perfluorodecalin (PFDEC) (n = 6) (both PFCs from F2 Chemical, Lancashire, UK). Because the two liquids used have the same density, heat capacity, and ventilation parameters (Table 1), both TLV groups were considered as only one group (n = 11) for the purposes of this article.

Table 1:
Ventilation Parameters Used During TLV for PFOB and PFDEC Liquids

Respiratory Management.

Total liquid ventilation was performed using a volume-controlled, pressure-limited, total liquid ventilator using independent piston pumps. This device is described in detail elsewhere.11 To initiate TLV, 10-ml aliquots of warmed (39°C), preoxygenated PFC (either PFOB or PFDEC) were sequentially instilled using the filling sequence of the TLV ventilator, during CMV, through the side port of the liquid ventilator connecting piece. The volumes of the aliquots were adjusted to achieve lambs' calculated functional residual capacity. On completion in which a total of 25 ml/kg of PFC was instilled in no more than 15 minutes, the animal was then shifted in TLV in horizontal position at the following initial ventilator settings: rate = 3–5 breaths/min with I:E = 1:3; end-inspiratory and end-expiratory pause = 0.5 seconds; Vt = 20 ml/kg; PEEP = 5 cm H2O, and FiO2 = 1.0. An exponential profile was used during expiration and a ramp profile during inspiration.

Conventional mechanical ventilation was performed using pressure-regulated, volume-controlled ventilation mode (Servo 300 ventilator, Siemens-Elema AB, Solna, Sweden) at the following initial ventilator settings: rate = 60 breaths/min with I:E = 1:2; Vt = 10 ml/kg; PEEP = 9 cm H2O, and FiO2 = 1.0.

Throughout the 240 minutes of experimentation, ventilator settings of both TLV and CMV were adjusted to target a PaO2 ≥100 Torr and a PaCO2 between 30 and 50 Torr.

Other Experimental Management.

Sodium bicarbonate or tromethamine was used to maintain pH >7.25. Infusion of crystalloids (bolus of 10 ml/kg normal saline or lactated Ringer's solution) or vasopressor (dopamine 2–20 μg/kg/min) was used as needed to maintain a mean arterial pressure ≥50 mm Hg. The rate of iv dextrose infusion was adjusted to maintain glucose blood levels at 40–100 mg/dl. At the end of the protocol, the animals were euthanized with a lethal dose of pentobarbital (60 mg/kg), and the lung and thoracic cavity were carefully inspected for evidence of perfluorothorax or gross abnormalities.

Transpulmonary and Pulmonary Thermodilution Measurement of Cardiac Output.

COTPTD and COPTD were measured using the above-described femoral and pulmonary artery thermodilution catheters, respectively, both connected to a commercially available device (PiCCOplus, PC8100, and VoLEF, PC8200, Pulsion Medical System, Munich, Germany). Measurements were stored using acquisition software specially designed by our laboratory (using data structure provided by Pulsion Medical System), which allowed simultaneous calculation of COTPTD and COPTD using the modified Stewart-Hamilton equation.12 Thermodilution measurements were performed at 30-minute intervals for the 240 minutes of experimentation with the first measurement made 30 minutes after randomization. A 5-ml bolus of ice-cold saline (<8°C) was manually injected into the jugular vein to perform all thermodilution measurements. At each time point, the mean CO of three consecutive thermodilution tests was used to compare the difference between COTPTD and COPTD. Thermodilutions were performed according to PiCCO and VoLEF manufacturer's recommendations, and results were standardized as CI using body surface area calculated according the following formula: weight0.67 × 0.085.13

Statistical Analysis

Intrainstrument Reproducibility of TPTD and PTD.

The level of precision for CO measurements between repeated measurements by TPTD and PTD were computed using the intrainstrument intraclass correlation coefficient (ICC).14 An ICC value of 1 indicates perfect agreement, with random or systematic differences between the measurements decreasing the value. An ICC value of up to 0.40 was considered to indicate positive but poor repeatability; 0.41–0.60 good repeatability; 0.61–0.80 very good repeatability; and >0.80 excellent repeatability.15 One-sided 95% confidence intervals (95% CI) are reported for all intrainstrument ICC.

Interchangeability of TPTD and PTD.

Linear regression and Bland-Altman analysis16 were both used to ascertain the interchangeability of TPTD with PTD. In Bland-Altman analysis, biases were computed as the mean differences between measurements of TPTD and PTD and subsequently plotted against the mean TPTD and PTD CI measurements. Limits of agreement (LOA) on Bland-Altman plots were calculated as ±1.96 × SD of the bias for the mean of three measurements17 and used to assess the range of agreement between the techniques. The range for acceptable LOA between both techniques of thermodilution was fixed to a percentage error [100% × (1.96 × SD of the bias)/mean CO] of ±28.4%, as suggested by Critchley and Critchley.18 The LOA also enables to assess the variability of the bias between the methods whereas the disposition of points on the Bland-Altman plots allows to visualize changes in method agreement according to the extent of the mean CI.16,17 A linear regression model was used with PTD as the dependent variable and TPTD as the independent variable. The 95% confidence intervals for the slope and intercept are reported. When the slope of the line is close to unity and the intercept close to zero, this implies that CI measured by both methods is similar. Finally, a parametric mixed linear model (mixed procedure) of SAS system (SAS Institute, Inc., v.9.1.3, Cary, NC) was used to assess the statistical significance (≤0.05) of the difference between the bias found in TLV compared with CMV groups.

Linear regressions were performed using SPSS statistics software v.17.0 for Macintosh (SPSS, Chicago, IL), and Bland-Altman graphs were plotted using Excel for Macintosh (Office 2008, Microsoft, Seattle, WA). All data are presented as mean ± SD.


A total of 71 and 45 pairs of cardiac indices were successfully collected in TLV and CMV, respectively. Mean CITPTD and mean CIPTD are provided, and other major hemodynamic parameters are presented in Tables 2 and 3, respectively. Mean transit times (MTt) of the thermal indicator during TPTD tests were similar with both perfluorochemical liquids but 36% higher than CMV group (Table 4).

Table 2:
Comparison of the Mean Cardiac Index Measured by Transpulmonary Thermodilution and Pulmonary Thermodilution as well as the Mean Absolute Bias and Limits of Agreement of the Bias in Both Conventional Mechanical Ventilation and Total Liquid Ventilation
Table 3:
Major Hemodynamic Parameters Pre- and Post-Thermodilution Tests
Table 4:
Mean Transit Time for the Different Types of Mechanical Ventilation

Thermodilution Reproducibility

Intrainstrument reproducibility for CI measurements of both devices using triplicate measurements was excellent for the group of lambs under CMV and very good for the group under TLV. ICC values of TPTD and PTD with one-sided confidence 95% interval (lower bound) were 0.990 (0.985) and 0.992 (0.988) in CMV and 0.686 (0.594) and 0.851 (0.799) in TLV, respectively.

Interchangeability of TPTD and PTD

Interchangeability of TPTD with PTD under both modalities of ventilation was poor. As depicted on the Bland-Altman plots (Figure 1, A and B), a bias of 0.98 L/min/m2 (17.3%) with wide LOA of −1.33 to 3.25 L/min/m2 was found in CMV compared with 1.28 L/min/m2 (22.8%) with an LOA of −1.17 to 3.72 L/min/m2 in TLV. Percentage errors were 44.8% and 48.7% in CMV and TLV groups, respectively. Linear regressions also indicated that the estimate of the intercept was close to zero, whereas the estimate of the slope was different from unity. In the CMV group, intercept was 0.006 (95% CI −0.770 to 0.781) and slope 0.826 (95% CI 0.696–0.957), whereas intercept and slope were −0.232 (95% CI −1.067 to 0.602) and 0.814 (95% CI 0.669–0.959) in the TLV group, respectively (Figure 2, A and B). Finally, biases were similar within both groups (p = 0.11).

Figure 1.:
Bland-Altman plots depicting mean bias and range of agreement between transpulmonary thermodilution cardiac index (CITPTD) and pulmonary thermodilution cardiac index (CIPTD) measurements in the (A) conventional mechanical ventilation (CMV) and (B) total liquid ventilation (TLV) groups.
Figure 2.:
Linear regression with intercept and slope (see equation at the bottom) and 95% confidence intervals of the mean (inner) and for individual measurements (outer), depicting the agreement between transpulmonary thermodilution cardiac index and pulmonary thermodilution measurements in the (A) conventional mechanical ventilation (CMV) and (B) total liquid ventilation (TLV) groups.


In pediatric and neonatal patients, TPTD has been validated by comparing this technique to Fick's method,1,3,4 ultrasonic flow probes6 and pulmonary thermodilution.3,19,20 These studies have notably led to the establishment of TPTD as the new “gold standard” for pediatric CI measurement under gas ventilation.5 In this study, intrainstrument ICC revealed highly reproducible measurements of CITPTD and CIPTD irrespective of the study group, thus indicating that TPTD and PTD remained precise techniques even under TLV. However, one may note that reproducibility of both thermodilution techniques is greater in CMV than TLV, and the precision of PTD seemed superior than TPTD in TLV.

Contrary to previous investigations on sheep,6,8,9 both thermodilution techniques seem to be neither interchangeable during CMV nor during TLV in our study. The LOA on Bland-Altman plots were widely over the range of acceptability of 28.3% defined by Critchley and Critchley18 in both groups, suggesting that TPTD and PTD are not interchangeable. Linear regressions showed similar results with slopes different from unity and a large 95% confidence interval for both TLV and CMV. Interestingly, we found no statistically significant difference between biases in CMV and TLV (p = 0.11). This observation suggests that PFC-filled lungs did not contribute significantly to the bias between TPTD. Consequently, it is rational to claim that diffusive losses of thermal indicator, if existing, were minimal in our study and did not lead to artifactual increase of CIs. Otherwise, one would expect the discrepancy between TPTD and PTD to be significantly higher in TLV than in CMV. Nevertheless, one can notice that MTt are significantly higher in TLV group depicting the greater intrathoracic thermal volume attributable to the PFC-filled lung.

The relative bias computed between both thermodilution methods was 22.8% and 17.3% in CMV and TLV, respectively. Although no comparative study exists in TLV, the bias found in the CMV group is slightly higher than that previously reported in animals (14.5% and 9.5% in piglets weighting on average 14 kg)19,20 as well as in young children (4.5% with median weight of 14 kg).3 Various factors other than the animal model itself may explain the biases in both groups, including the slowing of sinusal rhythm after injection of the ice-cold bolus.21–23 In our study, thermodilution tests induced highly variable but systematic heart rate slowdown (Table 3) in which 24% occasioned significant bradycardia (>10%) lasting on average 24.3 ± 7.8 seconds (data not shown). The femoral catheter took advantage of a longer period of time between injection and the sensing of the thermal indicator, which attenuates the effects of the transient bradycardia on TPTD.20 The heart rate slowdown observed could therefore preferentially underestimate CI as measured by PTD and contribute to increase in the biases. On the other hand, given the animal size, this effect could be partly compensated by the increase of CO after the triplicate injections of cold bolus. However, there were no significant variations in major hemodynamic parameters pre- and post-thermodilution supporting this issue in our study (Table 3). Central venous pressure increased slightly post-thermodilution compared with pre-thermodilution value, but the variability of these measurements was too high to interpret them.

Study Limitations

Extension of these results to infants in intensive care setting should be done with caution because thermodilution measurements were performed in an animal model with limited number of data. Because no significant differences in major ventilatory (Table 1) and cardiovascular10 parameters were found between PFOB and PFDEC, both perfluorochemical liquids have been gathered to increase the number of data in the TLV group, despite introducing some heterogeneity in this group.


We achieved the first comparison of TPTD and PTD in TLV and report herein that both thermodilution methods stay highly precise techniques even in PFC-filled lungs. Despite significant biases between both thermodilution techniques, TLV did not bring additional bias compared with CMV group in our study. This finding suggests that TLV does not interfere with TPTD measurement. Because the later is less invasive, we advocate the use of TPTD for hemodynamic monitoring as recommended in CMV.


The authors thank Pulsion Medical Systems AG that provided the PiCCO and VoLEF devices.

Supported by the Fonds de Recherche sur la Nature et les Technologies, the Foundation of Stars, and l'Université de Sherbrooke.


1. Tibby SM, Hatherill M, Marsh MJ, et al: Clinical validation of cardiac output measurements using femoral artery thermodilution with direct Fick in ventilated children and infants. Intensive Care Med 23: 987–991, 1997.
2. Frazier SK, Skinner GJ: Pulmonary artery catheters: State of the controversy. J Cardiovasc Nurs 23: 113–121, 2008.
3. McLuckie A, Murdoch IA, Marsh MJ, Anderson D: A comparison of pulmonary and femoral artery thermodilution cardiac indices in paediatric intensive care patients. Acta Paediatr 85: 336–338, 1996.
4. Pauli C, Fakler U, Genz T, et al: Cardiac output determination in children: Equivalence of the transpulmonary thermodilution method to the direct Fick principle. Intensive Care Med 28: 947–952, 2002.
5. Tibby S: Transpulmonary thermodilution: Finally, a gold standard for pediatric cardiac output measurement. Pediatr Crit Care Med 9: 341–342, 2008.
6. Lemson J, de Boode WP, Hopman JC, et al: Validation of transpulmonary thermodilution cardiac output measurement in a pediatric animal model. Pediatr Crit Care Med 9: 313–319, 2008.
7. Wolfson MR, Shaffer TH: Pulmonary applications of perfluorochemical liquids: Ventilation and beyond. Paediatr Respir Rev 6: 117–127, 2005.
8. Fessler HE, Pearse D: Accuracy of hemodynamic measurements during partial liquid ventilation with perflubron. Am J Respir Crit Care Med 162: 1372–1376, 2000.
9. Meinhardt JP, Friess U, Bender HJ, et al: Relationship among cardiac index, inspiration/expiration ratio, and perfluorocarbon dose during partial liquid ventilation in an oleic acid model of acute lung injury in sheep. J Pediatr Surg 40: 1395–1403, 2005.
10. Walti H, Lebon J, Robert R, et al: Efficacy and cardiovascular consequences of total liquid ventilation (TLV) in experimental respiratory distress syndrome (ARDS). Arch Dis Child 93: 500, 2008.
11. Robert R, Micheau P, Cyr S, et al: A prototype of volume-controlled tidal liquid ventilator using independent piston pumps. ASAIO J 52: 638–645, 2006.
12. Isakow W, Schuster DP: Extravascular lung water measurements and hemodynamic monitoring in the critically ill: Bedside alternatives to the pulmonary artery catheter. Am J Physiol Lung Cell Mol Physiol 291: L1118–L1131, 2006.
13. Mitchell HH: Check formulas for surface area of sheep, in A Year's Progress in Solving Farm Problems in Illinois. Illinois Agricultural Experimental Station Annual Report. Champaign, IL, 1928.
14. Shrout PE, Fleiss JL: Intraclass correlations: Uses in assessing rater reliability. Psychol Bull 86: 420–428, 1979.
15. Landis JR, Koch GG: The measurement of observer agreement for categorical data. Biometrics 33: 159–174, 1977.
16. Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1: 307–310, 1986.
17. Bland JM, Altman DG: Measuring agreement in method comparison studies. Stat Methods Med Res 8: 135–160, 1999.
18. Critchley LA, Critchley JA: A meta-analysis of studies using bias and precision statistics to compare cardiac output measurement techniques. J Clin Monit Comput 15: 85–91, 1999.
19. Marx G, Sumpelmann R, Schuerholz T, et al: Cardiac output measurement by arterial thermodilution in piglets. Anesth Analg 90: 57–58, 2000.
20. Ruperez M, Lopez-Herce J, Garcia C, et al: Comparison between cardiac output measured by the pulmonary arterial thermodilution technique and that measured by the femoral arterial thermodilution technique in a pediatric animal model. Pediatr Cardiol 25: 119–123, 2004.
21. Harris AP, Miller CF, Beattie C, et al: The slowing of sinus rhythm during thermodilution cardiac output determination and the effect of altering injectate temperature. Anesthesiology 63: 540–541, 1985.
22. Marshall JM: Effects of low temperatures on transmembrane potentials of single fibers of the rabbit atrium. Circ Res 5: 664–669, 1957.
23. Nishikawa T, Namiki A: Mechanism for slowing of heart rate and associated changes in pulmonary circulation elicited by cold injectate during thermodilution cardiac output determination in dogs. Anesthesiology 68: 221–225, 1988.
24. Nadeau S, Noble WH: Limitations of cardiac output measurements by thermodilution. Can Anaesth Soc J 33: 780–784, 1986.
    25. Nishikawa T, Dohi S: Errors in the measurement of cardiac output by thermodilution. Can J Anaesth 40: 142–153, 1993.
      26. van Grondelle A, Ditchey RV, Groves BM, et al: Thermodilution method overestimates low cardiac output in humans. Am J Physiol 245: H690–H692, 1983.
        Copyright © 2010 by the American Society for Artificial Internal Organs