The history of studying human circulation is a complex one, with various contributions, including those from physicians, artists, and even theologians. Knowledge about the components of the cardiovascular system dates back to the concept of kardia established by the Greeks and Aristotle’s assertion that the heart had a role in generating body heat. Around 200 AD, Galen proposed the idea that the liver was responsible for converting chyle into blood and distributing it by means of the veins to the rest of the body, including the heart. Though held for years, this belief did not incorporate the concept of circulation, and by the middle ages doubts arose regarding Galen’s idea. Leonardo da Vinci subsequently advanced our understanding of circulation through his detailed descriptions of the ox heart and its 4 chambers, as well as the function of great vessel valves in creating unidirectional flow. However, at this time, there was still no reasonable explanation for how blood passed from the right to the left ventricle. By the late 1500s, Vesalius addressed this issue in his book, The Fabric of the Human Body, and the theological works of Servetus introduced the concept of a lesser circulation that involved the passage of blood from the right to the left side of the heart. With Fabricius’s description of circulation in his book, On the Valves of the Veins in 1574, knowledge of the cardiovascular system began to broaden.1 Harvey eventually established the basic and descriptive concepts of circulation in his lecture notes from Prelectiones Anatomiae Universalis in 1616.2 This would later be followed by his more descriptive piece of work, An Anatomical Dissertation on the Motion of the Heart and Blood in Animals, in 1628. These contributions helped establish the concept of circulating blood; however, it would still be centuries before Fick and Stewart began their work of quantifying blood flow.
This article will review thermodilution cardiac output (CO) as part of physiologic hemodynamic measurement.
MEASUREMENT OF CARDIAC OUTPUT
At the time of its creation by Adolf Fick in 1870, CO measurement, defined as the amount of liters pumped by the heart in minutes, was only a theoretical principle that could not yet be performed.3 The Fick principle reflects the extraction of oxygen through the systemic circulation and is dependent on arterial and pulmonary arterial oxygen levels, hemoglobin, and the maximum amount of oxygen consumption by the body in a given period (VO2 max). Thus, the Fick principle involves the blood oxygenation process through the lungs and oxygen delivery to the tissues, where it is added and removed by a flow limited mechanism. By measuring the amount of oxygen carried by the blood, the total amount of blood needed to carry oxygen through the lungs could be calculated. This value then reflects the output of the heart.
According to Stewart’s discussion on indicator techniques in 1893, Haller had previously used a colored liquid in euthanized animals to evaluate blood circulation through the lungs in 1761.4 Stewart’s method, however, differed from this indicator approach. It consisted of using electrolyte solutions as indicators in a process similar to, but more refined than, the technique described by Hering in 1827.4,5 Stewart’s initial experiments on blood circulation time were done by injecting sodium chloride solution into animals and recording the time required for the indicator to reach a specific vessel. These studies were intended to determine factors that increased or diminished circulation to organs. Vessel indicator detection was achieved through electrodes connected to a Wheatstone bridge that signaled changes in blood resistance caused by circulating sodium chloride. The measurements provided the time of indicator circulation from point of injection to point of detection.
Thirty years after his initial studies, Stewart improved previous indicator techniques to obtain CO measurements in animals by using a sodium chloride indicator and taking blood samples.6,7 This allowed the quantification of blood volume by determining the extent of sodium chloride diluted. Using the same sodium chloride indicator detection system, Stewart drew blood samples from the vessel contralateral to the one connected to electrodes whenever the indicator circulated through the detection site and produced changes in resistance. The resulting dilution rate was used in an algebraic method representing blood flow based on indicator concentration change to provide these initial CO measurements in dogs.
Further improvements in indicator techniques came in subsequent years. Hamilton, an important contributor to the subject, used measurable dye indicators to calculate CO in humans in the 1930’s.8–10 The procedure involved the injection of dye into the venous circulation, with simultaneous insertion of an arterial catheter into the radial artery for continuous blood samples measuring indicator concentration. Concentration changes over time were plotted as curves that would then be used for CO calculation. Though this technique would later be improved, these initial studies paved the way to what is now known as the Stewart-Hamilton CO equation. Thermodilution somewhat resembles these techniques but provides comparable measurements through a less complex indicator detection mechanism.
MEASUREMENT OF CARDIAC OUTPUT BY THERMODILUTION
The theoretical basis for thermodilution was introduced in 1953 by Fegler, who began performing animal experiments to measure CO a year later.11 He conducted testing on dogs using cold Ringer solution as an indicator. Two constantan-copper thermocouples inserted through the right jugular vein into the right ventricle and from the left carotid or right femoral artery into the aorta, generated intrinsic temperature-dependent voltage to measure right- and left-sided blood temperature changes that were used to calculate CO.12 Fluoroscopy was used to keep the catheters in place, and a Tinsley galvanometer connected to the thermocouples recorded blood temperature changes. Values and CO calculations resembled those of dye dilution techniques. The accuracy of this technique was established by comparing results to those from in vitro flow measurements and the Fick method and demonstrating comparably significant values for flow and CO, respectively.11
Early criticism of thermodilution drove subsequent testing to further establish and validate its meaningful use in single vessel blood flow measurements.13 Ganz et al14 simplified the thermodilution CO method in 1971. The new and improved technique consisted of using 2 catheters, a Lehman and a Teflon, with the Lehman placed in the superior vena cava as the injection catheter and the Teflon passed through the lumen of the Lehman into the pulmonary artery (PA). The Lehman was equipped with 2 thermistors, small semiconductors (Fig. 1) made from metals or polymers that work as thermally-sensitive resistors and temperature-sensing devices.12,15 The proximal thermistor remained in the superior vena cava and measured the indicator temperature as it left the catheter. The distal thermistor detected temperature changes induced by the indicator in the PA. The PA measurements helped avoid significant indicator recirculation.
In 1972, Forrester et al16 used an updated balloon-tipped flow-directed thermodilution PA catheter (PAC). The initial indicator temperature could be measured outside the body, and the proximal thermistor was therefore abandoned. The balloon tip guided and localized the distal catheter with its thermistor in the PA. Cold normal saline was injected into the right atrium through the balloon-tipped catheter, and temperature measurements were made in the PA. In conjunction with newly developed analogue computers used to calculate CO, this technique made thermodilution a bedside procedure.17–20 The standard thermodilution PAC (Fig. 2) has 4 lumens serving 4 corresponding functions: balloon inflation, indicator injection, solution injection or pressure monitoring, and achievement of the thermistor circuit.
Indicator Dilution Techniques
The volume of a stationary liquid can be quantified directly or indirectly. The indirect approach resembles those of indicator dilution techniques and will be discussed here. Suppose a 1-mL solution containing 1 mg of a measurable substance (the indicator) is added to an unknown volume of liquid. If the new indicator concentration is measured, it can be used to determine the unknown volume. A new concentration of 0.05 mg/mL, eg, reveals that the indicator has been diluted 20 times and that the unknown volume in question is 19 mL. This is a simple way of understanding the ideas that underlie indicator dilution techniques for the measurement of blood flow.
The preceding example shows how to calculate the volume of a stationary liquid, but with adequate measurements, this concept can also be applied to a flowing liquid in a closed circuit. It is crucial to understand that as the indicator is added, its total volume is distributed across the flowing fluid, from initial time of administration to completion of the procedure. Even though the mixture may be uniform, the amount of indicator added is dispersed through the flowing liquid, and constant measurements are required to detect total circulating fluid volume and the new indicator concentration. The total volume of liquid that dilutes the indicator, along with the time interval from initial indicator detection time (t0) to its final detection time (tf), is necessary to determine the flow rate of the liquid. By way of example, suppose 1 mL containing 300 mg of an indicator is added to a flowing liquid. The indicator concentration is then measured in this liquid at 5-second time intervals from time of initial detection up to 30 seconds. The Figure 3 shows the indicator concentration measurements over time connected by a treadline and represents how the concentration changes as it is measured.
Note that the final measurement is obtained as the treadline nears the baseline. This is done to avoid remeasuring the indicator in the now closed circuit. The concentration changes over time represent the new total concentration in the flowing liquid. A simple way to approximate this concentration is to average all of the measurements, which in this example produces a value of 11.85 mg/100 mL. To determine the total volume of flowing liquid necessary to achieve this indicator concentration, a calculation similar to the one previously mentioned is then performed:
Initial Indicator Concentration: 300 mg/mL
Since the indicator was diluted over a 30-second time period, the volume of liquid moved per unit of time, or flow, can be calculated in liters per minute (L/min).
The liquid is therefore flowing at approximately 5 L/min. Notice that it is not necessary to subtract the volume of indicator added as it is incorporated into the total flow volume. The method described in this example, which resembles initial CO experiments performed by Hamilton in which the indicator was injected into the jugular vein and measured from radial artery samples, provides the means to measure flow through the heart.8 The accepted procedure, however, is more complex, with a continuum of concentration measurements and curve integration providing better representation of indicator circulation. The simplified Stewart-Hamilton equation used to determine CO is as follows:
In this formula, the numerator is the initial indicator added. The denominator represents the indicator concentration change (final concentration) over a given time period and can be calculated through integration or determination of the area under the concentration-time curve.
Thermodilution is similar to this earlier procedure, but it is performed by injecting a cold or room temperature physiologic solution into the right atrium and directly measuring induced temperature changes afterwards in the PA. The recorded temperature-time curve is representative of a concentration-time curve, and the area under the curve reflects the amount of indicator carried by a volume of blood per unit of time.
THE THERMODILUTION METHOD
The Stewart-Hamilton equation shows the relationship between blood flow, the indicator, and its blood concentration. However, an adjustment is necessary if the equation is used for thermodilution because temperature changes, rather than indicator concentrations, are measured. The change in temperature represents an unknown mass of blood losing heat to a known mass of cold indicator, and it is therefore possible to calculate blood volume by using temperature change measurements to calculate the blood mass (in grams) and convert that value into blood volume. This is done by incorporating 2 important variables for each of the 2 fluids—specific heat and specific gravity—to the equation. The specific heat represents the energy needed to change the temperature of 1 g of a substance by 1°C, and the specific gravity is the density of a substance in relation to the density of water. The units for specific heat and specific gravity are cal/°C·g and mg/mL, respectively. For the indicator, the volume added and initial temperature are both known, yielding the following:
From equations 3 and 4, the following relationship can be obtained:
The preceding simplifications show how it is possible to determine the unknown blood volume from measured temperature changes using the ratio between indicator and blood properties. The specific heat and gravity values of fluids used are listed in the Table 1.13,14
The blood temperature changes measured by thermodilution are graphed on a temperature-time curve. The original curve shows a sharp negative deflection followed by a rise to an apex and gradual return to baseline values. This trend represents the initial drop in blood temperature followed by the increase observed as the blood dilutes and carries out the cold indicator. For easier representation, the thermodilution curves are placed as upright deflections (Fig. 4), and the area under the curve, which represents the indicator concentration as a function of time, is inversely proportional to CO.21
In the indicator dilution method, CO is calculated using the Stewart-Hamilton equation, described previously in simplified form (Equation 2). The equation is based on the law of conservation of mass, where the dilution rate, or concentration change in the indicator added to a moving liquid, is used to calculate flow.22 Thus, several ideal conditions are assumed, including single inflow and outflow tracts, no indicator loss, complete mixing of the indicator, constant blood and indicator flow rate, and no indicator recirculation.23–25 For thermodilution, it is necessary that the temperature changes reflect indicator concentration as blood flows, which is achievable with the specific heat and specific gravity for both indicator and blood. Additional factors that correct for indicator warming are also added to the equation. The resulting modified Stewart-Hamilton equation used to calculate CO by thermodilution is as follows21:
The absolute indicator volume injected (V) depends directly on the difference between the initial blood (TB) and indicator (TI) temperatures.23 The denominator represents the area (A) under the thermodilution curve that reflects the change in temperature as blood flows. This value is calculated using a computer connected to the thermistor and divided by the graph recording speed (mm/s). The density factor (K1) is calculated with the specific heat and specific gravity of the indicator and blood. The calibration constant (K2) represents the temperature change (°C) corresponding to 1 mm of curve amplitude.13 The value of 60 is used to convert mL/s into mL/min. The correction factor (C) is empirical and provided by the thermodilution system manufacturer. It accounts for the amount of indicator left in the catheter, heat change during injection, injection rate, and indicator loss during the procedure.21,26–28
CONTINUOUS CARDIAC OUTPUT MEASUREMENT
In in vitro settings, continuous indicator flow measurement demonstrates less variability than pulsatile flow measurements, and new techniques have been aimed at achieving a continuous CO reading.29 The systems for monitoring provide a constant CO readout, much like continuous heart rate and oximetry in bedside monitors. In the early 1960’s, a technique was tested using a catheter with a high frequency electric heater introduced into the right atrium and superior vena cava.30 The heater warmed at a predetermined rate for a short period, and temperature measurements were made simultaneously in the PA. Using this same principle, a coronary sinus flow meter was developed, and a similar mechanism introduced by Yelderman in 1990 was then used to measure CO with a stochastic heating system.31,32 Yelderman’s technique uses a PAC mounted with a 10-cm filament proximal to the thermistor, and temperature changes are continuously measured in the PA as the filament warms at random intervals and heats the surrounding environment in the right ventricle. An average of the recorded measurements is then used to generate continuous CO values in a process that has proven so reliable compared with standard pulsatile or bolus techniques that it has been accepted in the critical care setting.33–36
LIMITATIONS TO THERMODILUTION CARDIAC OUTPUT
Thermodilution CO measurements are affected by various disease states and anatomic defects. When CO is low, thermodilution becomes less reliable because of the small temperature change induced as the indicator warms up in the diminished circulation.37–39 The smaller area under the curve causes an overestimated CO. A similar effect is seen in the setting of severe tricuspid regurgitation where thermodilution yields lower accuracy compared with the Fick method.40,41 Underestimation of CO can also result from high flow states that cause thermodilution variability that is potentially related to rapid PA temperature variation and limited thermistor sensitivity to rapid temperature change.42 Pulmonary valve insufficiency has been shown to affect the appearance of the thermodilution curve, but it does not affect its overall value.43
Although CO extremes cause thermodilution measurement variability, values within a certain range are useful. Thermodilution-measured CO values between 3.4 and 15.8 L/min have a correlation coefficient of 0.98 compared with dye dilution methods.44 Comparable results between Fick and thermodilution are observed in CO from 3.2 to 17.5 L/min.45 Only in values >15 L/min does minimal variability appear, and thermodilution begins to underestimate CO.42
CARDIAC OUTPUT AND ITS USE
CO measurements by a PAC are used predominantly in the critical care setting as invasive cardiac function parameters that allow for both intracardiac pressure and CO monitoring. The physiologic information obtained is diagnostically valuable in titrating and guiding pharmacotherapy and fluid administration. CO values can also be used to evaluate acute heart failure through Forrester hemodynamic subsets, which classify patients into 4 groups based on cardiac index and pulmonary wedge pressure after acute myocardial infarction.46 By correlating these values with patient mortality, the subsets serve as a guide in treating patients in each of the 4 classification groups.
In the last decade, data obtained from the Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness (ESCAPE trial) and the Acute Respiratory Distress Syndrome (ARDS) clinical trials network suggest that the PAC does not improve patient mortality or hospital stay and may even be associated with more complications.47,48 A meta-analysis of previously done clinical randomized trials produced similar findings.49 Collectively, these studies have cast doubt on invasive diagnostic approaches in assessing heart function and reveal that cardiac parameter-guided therapy has not translated into significant reductions in critically-ill patient mortality. The resulting reduction in use of the PAC in medical and surgical patients may in turn reflect a reduced need for CO measurements.50 It is important to mention that in these previous studies there was no indication of who interpreted the clinical data obtained from the measurements and to what end this was used.
In the current era, there continues to be a role for the concept of CO and PAC use. These concepts are valuable when considering mechanical circulatory support and cardiac transplantation. They provide an objective diagnostic assessment of the systemic and pulmonary hemodynamic clinical status. With the information obtained (eg, intracardiac pressures, CO, and derived parameters, such as vascular resistance), patients are directed into different treatment modalities. With the availability of advanced and invasive treatment strategies for conditions such as heart failure and pulmonary hypertension, there continues to be a role for PAC and CO measurement.
The ideas and concepts that underlie thermodilution date back over 100 years. Knowledge of the many mechanical and chemical aspects of science can lead to everyday innovations, and thermodilution CO is a clear example of how physics and medicine have intersected to produce important, new knowledge. By always recognizing the science behind such technology, the medical community will be able to produce new advancements that meet the emerging needs of the health care profession. Ultimately, it is interesting, humbling, and instructive to note that despite the many years it took to reliably measure something as basic as CO, its use continues to evolve, aiming to fully benefit those who need it the most.
1. Opie LH. Heart Physiology: From Cell to Circulation. 2004.4th ed. Philadelphia, PA: Lippincott Williams & Wilkins.
2. Harvey W. Leaky CD. Exercitatio Anatomica de Motu Cordis et Sanguinis in Animalibus, english translation by 1930:3rd ed. Springfield, IL: Charles C Thomas; 150.
3. Grossman W, Baim DS. Grossman’s Cardiac Catheterization, Angiography, and Intervention. 2000.6th ed. Philadelphia, PA: Lippincott Wiliam & Wilkins.
4. Stewart GN. Researches on the circulation time in organs and on the influences which affect it: parts I.-III. J Physiol. 1893;15:1–89.
5. Fox IJ. History and developmental aspects of the indicator-dilution technique. Circ Res. 1962;10:381–392.
6. Stewart GN. Researches on the circulation time and on the influences which affect it. J Physiol. 1897;22:159–183.
7. Stewart GN. The output of the heart in dogs. Am J of Physiol. 1921;57:27–50.
8. Hamilton WF, Moore JW, Kinsman JM, et al. Simultaneous determination of the pulmonary and systemic circulation times in man and of a figure related to the cardiac output. Am J of Physiol. 1928;84:338–334.
9. Moore JW, Kinsman JM, Hamilton WF, et al. Studies on the circulation II: cardiac output determinations; comparison of the injection method with the direct Fick procedure. Am J of Physiol. 1929;89:331–339.
10. Hamilton WF, Moore JW, Kinsman JM, et al. Studies on the circulation IV: further analysis of the injection method, and of changes in hemodynamics
under physiological and pathological conditions. Am J of Physiol. 1932;99:534–551.
11. Fegler G. Measurement of cardiac output in anaesthetized animals by a thermodilution method. Q J Exp Physiol Cogn Med Sci. 1954;39:153–164.
12. Wheeler A, Ganji A. Introduction to Engineering Experimentation. 2003.2nd ed. Upper Saddle River, NJ: Prentice Hall.
13. Fronek A, Ganz V. Measurement of flow in single blood vessels including cardiac output by local thermodilution. Quart J Exp Physiol. 1960;8:175–182.
14. Ganz W, Donoso R, Marcus HS, et al. A new technique for measurement of cardiac output by thermodilution in man. Am J Cardiol. 1971;27:392–396.
15. Profio AE. Biomedical Engineering. 1993.New York, NY: Wiley.
16. Forrester JS, Ganz W, Diamond G, et al. Thermodilution cardiac output determination with a single flow-directed catheter. Am Heart J. 1972;83:306–311.
17. Phillips CM, Davila JC, Sanmarco ME. Measurement of cardiac output by thermal dilution. II. A new computer for rapid, convenient determinations. Med Res Eng. 1970;9:25–29.
18. Olsson SB, Wassén R, Varnauskas E, et al. A simple analogue computer for cardiac output determination by thermodilution. Cardiovasc Res. 1972;6:303–308.
19. Ellis RJ, Gold J, Rees JR, et al. Computerized monitoring of cardiac output by thermal dilution. JAMA. 1972;220:507–511.
20. Hendriks FF, Schipperheyn JJ, Quanjer PH. Thermal dilution measurement of cardiac output in dogs using an analog computer. Basic Res Cardiol. 1978;73:459–468.
21. Headley J. Invasive Hemodynamic Monitoring: Physiological Principles and Clinical Applications. 2002:Irvine, CA: Edwards LifeSciences; 19–20.
22. Berne RM, Levy MN, Koeppen BM, et al. Physiology. 2004.5th ed. Philadelphia, PA: Elsevier Inc.
23. Zierler KL. Theoretical basis of indicator-dilution methods for measuring flow and volume. Circ Res. 1962;10:393–407.
24. Hosie KF. Thermal-dilution techniques. Circ Res. 1962;10:491–504.
25. Jansen JR, Schreuder JJ, Settels JJ, et al. An adequate strategy for the thermodilution technique in patients during mechanical ventilation. Intensive Care Med. 1990;16:422–425.
26. Levett JM, Replogle RL. Thermodilution cardiac output: a critical analysis and review of the literature. J Surg Res. 1979;27:392–404.
27. Weisel RD, Berger RL, Hechtman HB. Current concepts measurement of cardiac output by thermodilution. N Engl J Med. 1975;292:682–684.
28. Ganz W, Swan HJ. Measurement of blood flow by thermodilution. Am J Cardiol. 1972;29:241–246.
29. Mackenzie JD, Haites NE, Rawles JM. Method of assessing the reproducibility of blood flow measurement: factors influencing the performance of thermodilution cardiac output computers. Br Heart J. 1986;55:14–24.
30. Khalil HH. Determination of cardiac output in man by a new method based on thermodilution. Lancet. 1963;1:1352–1354.
31. Afonso S. A thermodilution flowmeter. J Appl Physiol. 1966;21:1883–1886.
32. Yelderman M. Continuous measurement of cardiac output with the use of stochastic system identification techniques. J Clin Monit. 1990;6:322–332.
33. Yelderman ML, Ramsay MA, Quinn MD, et al. Continuous thermodilution cardiac output measurement in intensive care unit patients. J Cardiothorac Vasc Anesth. 1992;6:270–274.
34. Siegel LC, Peearl RG. Comparison of cardiac output measurement using a heat exchange catheter versus thermodilution. Anesthesiology. 1994;81:A512.
35. Boldt J, Menges T, Wollbrück M, et al. Is continuous cardiac output measurement using thermodilution reliable in the critically ill patient? Crit Care Med. 1994;22:1913–1918.
36. Lefrant JY, Bruelle P, Ripart J, et al. Cardiac output measurement in critically ill patients: comparison of continuous and conventional thermodilution techniques. Can J Anaesth. 1995;42:972–976.
37. Bilfinger TV, Lin CY, Anagnostopoulos CE. In vitro determination of accuracy of cardiac output measurements by thermal dilution. J Surg Res. 1982;33:409–414.
38. Norris SL, King EG, Grace M, et al. Thermodilution cardiac output–an in vitro model of low flow states. Crit Care Med. 1986;14:57–59.
39. Hillis LD, Firth BG, Winniford MD. Comparison of thermodilution and indocyanine green dye in low cardiac output or left-sided regurgitation. Am J Cardiol. 1986;57:1201–1202.
40. Lipkin DP, Poole-Wilson PA. Measurement of cardiac output during exercise by the thermodilution and direct Fick techniques in patients with chronic congestive heart failure. Am J Cardiol. 1985;56:321–324.
41. Hamilton MA, Stevenson LW, Woo M, et al. Effect of tricuspid regurgitation on the reliability of the thermodilution cardiac output technique in congestive heart failure. Am J Cardiol. 1989;64:945–948.
42. Kubo SH, Burchenal JE, Cody RJ. Comparison of direct Fick and thermodilution cardiac output techniques at high flow rates. Am J Cardiol. 1987;59:384–386.
43. Beyer J, Lamberti JJ, Replogle RL. Validity of thermodilution cardiac output determination: experimental studies with and without pulmonary insufficiency. J Surg Res. 1976;21:313–317.
44. Olsson B, Pool J, Vandermoten P, et al. Validity and reproducibility of determination of cardiac output by thermodilution in man. Cardiology. 1970;55:136–148.
45. Branthwaite MA, Bradley RD. Measurement of cardiac output by thermal dilution in man. J Appl Physiol. 1968;24:434–438.
46. Forrester JS, Diamond G, Chatterjee K, et al. Medical therapy of acute myocardial infarction by application of hemodynamic subsets (first of two parts). N Engl J Med. 1976;295:1356–1362.
47. The ESCAPE Investigators and ESCAPE Study Coordinators. Evaluation study of congestive heart failure and pulmonary artery catheterization effectiveness: the ESCAPE Trial. JAMA. 2005; 294:1625–1633.
48. Wheeler AP, Bernard GR, Thompson BT, et al.; National Heart, Lung, and Blood Institute Acute Respiratory Distress Syndrome (ARDS) Clinical Trials Network; Pulmonary-artery versus central venous catheter to guide treatment of acute lung injury. N Engl J Med. 2006; 354:2213–2224.
49. Shah MR, Hasselblad V, Stevenson LW, et al. Impact of the pulmonary artery catheter in critically ill patients: meta-analysis of randomized clinical trials. JAMA. 2005;294:1664–1670.
50. Wiener RS, Welch HG. Trends in the use of the pulmonary artery catheter in the United States. 1993–2004. JAMA. 2007; 298:423–429.