Comparison of Transpulmonary Thermodilution and Calibrated Pulse Contour Analysis with Pulmonary Artery Thermodilution Cardiac Output Measurements in Anesthetized Dogs

Abstract

Background

Transpulmonary thermodilution (TPTD_CO) and calibrated pulse contour analysis (PCA_CO) are alternatives to pulmonary artery thermodilution cardiac output (PATD_CO) measurement.

Hypothesis

Ten mL of ice‐cold thermal indicator (TI ₁₀) would improve the agreement and trending ability between TPTD_CO and PATD_CO compared to 5 mL of indicator (TI ₅) (Phase‐1). The agreement and TA between PCA_CO and PATD_CO would be poor during changes in systemic vascular resistance (SVR) (Phase‐2).

Animals

Eight clinically normal dogs (20.8–31.5 kg).

Methods

Prospective, experimental study. Simultaneous TPTD_CO and PATD_CO (averaged from 3 repetitions) using TI ₅ and TI ₁₀ were obtained during isoflurane anesthesia combined or not with remifentanil or dobutamine (Phase‐1). Triplicate PCA_CO and PATD_CO measurements were recorded during phenylephrine‐induced vasoconstriction and nitroprusside‐induced vasodilation (Phase‐2).

Results

Mean bias (limits of agreement: LOA) (L/min), percentage bias (PB), and percentage error (PE) were 0.62 (−0.11 to 1.35), 16%, and 19% for TI ₅; and 0.33 (−0.25 to 0.91), 9%, and 16% for TI ₁₀. Mean bias (LOA), PB, and PE were 0.22 (−0.63 to 1.07), 6%, and 23% during phenylephrine; and 2.12 (0.70–3.55), 43%, and 29% during nitroprusside. Mean angular bias (radial LOA) values were 2° (−10° to 14°) and −1° (−9° to 6°) for TI ₅ and TI ₁₀, respectively (Phase‐1), and 38° (5°–71°) (Phase‐2).

Conclusions and Clinical Importance

Although TI ₁₀ slightly improves the agreement and trending ability between TPTD_CO and PATD_CO in comparison to TI ₅, both volumes can be used for TPTD_CO in replacement of PATD_CO. Vasodilation worsens the agreement between PCA_CO and PATD_CO. Because of PCA_CO's poor agreement and trending ability with PATD_CO during SVR changes, this method has limited clinical application.

Abbreviations

AOA

accuracy of agreement

CO

cardiac output

CRI

constant rate infusion

CVP

central venous pressure

HR

heart rate

LOA

limits of agreement

MAP

mean arterial pressure

PATD_CO

pulmonary artery thermodilution cardiac output

PB

percentage bias

PCA_CO

calibrated pulse contour analysis cardiac output

PE

percentage error

POA_REFxREF

precision of agreement expected if the reference technique was compared to itself

POA_TESTxREF

precision of agreement of test versus reference methods

POM

precision of method

POM_REF

precision of reference method

POM_TEST

precision of test method

SVR

systemic vascular resistance

TPTD_CO

transpulmonary thermodilution cardiac output

ΔCO

delta changes in CO

ΔT_blood

delta changes in blood temperature

Cardiac output (CO) monitoring can play an important role during the decision‐making process in critically ill patients that require cardiovascular stabilization. Although pulmonary artery thermodilution cardiac output (PATD_CO) has been the clinical gold standard in human medicine since its introduction in the early 1970s,1, 2 placement of a pulmonary artery catheter is an invasive procedure that does not improve survival in critically ill patients and might result in rare (overall incidence 0.1%) but potentially fatal complications, including right ventricular rupture, knotting, and pulmonary artery rupture.3, 4 Although PATD_CO is frequently used in animal experimentation, its clinical application in veterinary medicine is restricted.

Transpulmonary thermodilution cardiac output (TPTD_CO) is an alternative to PATD_CO that has become increasingly popular in human medicine because it does not require pulmonary artery catheterization.2, 5, 6 For both techniques, changes in blood temperature (ΔT_blood) induced by the rapid injection of a thermal indicator (ice‐cold or room temperature solution) into the vena cava or right atrium are recorded downstream for CO calculation.2, 5, 6 In PATD_CO, the ΔT_blood is recorded by a thermistor located in the pulmonary artery, whereas in TPTD_CO the ΔT_blood is recorded further downstream by a thermistor located in a central artery (usually the femoral artery).2, 5, 6 While placing a femoral catheter produce some risks in veterinary medicine, it appears more attractive then placing a pulmonary artery catheter because it is does not require cardiac catheterization. However, the longer distance between the site of ice‐cold signal injection and the site for measuring ΔT_blood results in greater loss of thermal signal in comparison to PATD_CO and can lead to CO overestimation.2, 5, 6, 7 This source of error could be particularly important in large animal species and during low CO states. Significant CO overestimation might also occur with lower volumes of thermal indicator because of decreased signal‐to‐noise ratio. Although the agreement between PATD_CO and TPTD_CO has been studied in humans,7, 8, 9, 10, 11 pigs,12 calves,13 cats,14, 15 and dogs,16, 17 objective interpretation of these reports is difficult because of the absence of clearly defined criteria to determine the clinically acceptable bias and limits of agreement (LOA) between PATD_CO and TPTD_CO.

Calibrated pulse contour analysis (PCA_CO) provides continuous CO estimations based on beat‐to‐beat variations in stroke volume calculated from complex analysis of the systolic portion of the arterial pressure waveform.5, 18 After an initial calibration of the system with TPTD_CO, continuous PCA_CO values are provided by the same monitor. Although a good agreement between PCA_CO and thermodilution CO techniques has been reported in the literature,19, 20, 21, 22 several studies have shown wide LOA between these methods during hemodynamic changes induced by hemorrhage, vasopressor or vasodilator therapy, and volume replacement.23, 24, 25, 26 Inaccurate CO estimations by PCA_CO might be caused by changes in systemic vascular resistance (SVR) because modifications in vascular tone alter the shape of the arterial blood pressure wave.25

The first hypothesis of the present study was that the use of a higher volume of thermal indicator (10 mL of ice‐cold physiological saline) for TPTD_CO would result in better agreement and ability to track changes in the clinical gold standard (PATD_CO) in comparison to a lower volume of thermal indicator (5 mL of ice‐cold physiological saline). The second hypothesis was that PCA_CO would result in poor agreement and trending ability with PATD_CO during drug‐induced changes in SVR.

Material and Methods

Animals

This study was approved by the Institutional Animal Care Committee (protocol number 93/2013). Eight clinically normal adult (27–28 months old) English Pointer breed dogs (3 intact nonpregnant females and 5 intact males), weighing 20.8–31.5 kg, were enrolled in this study. Food but not water was withheld 12 hours prior to each experiment. Health status was evaluated by means of CBC, biochemical profile and venous blood gases that were within normal ranges.

Instrumentation

After placing a 20‐gauge catheter¹ in a cephalic vein, anesthesia was induced with intravenous propofol² (6.0 ± 1.0 mg/kg) and maintained with isoflurane.³ End‐tidal isoflurane concentrations were monitored⁴ and maintained at 1.5 times the minimum alveolar concentration (MAC) (1.78 ± 0.41% at sea level) throughout the study. Isoflurane MAC values were obtained for each individual animal in a preliminary study.

Animals were mechanically ventilated⁴ with an inspired oxygen fraction of 60% and received intravenous Lactated Ringer's solution (2 mL/kg/h) throughout anesthesia. Tidal volume was set at 12 mL/kg with an inspiration‐to‐expiration ratio of 1 : 2, whereas the respiratory rate was adjusted to maintain PaCO₂ between 35 and 45 mmHg.

A 22‐Gauge catheter¹, inserted percutaneously into the femoral artery 2.5 cm away from the inguinal fold, was used for the introduction of a wire that served as a guide for the subsequent insertion of a 3‐french, 7 cm long thermistor‐tipped catheter⁵ into the femoral artery based on the Seldinger technique. The pressure sensing lumen of the femoral artery catheter⁵ was connected to a pressure transducer⁶ via noncompliant tubing filled with heparinized (4 UI/mL) physiological saline. The transducer was zeroed at the level of the heart and connected to a monitor⁷ to display systolic, diastolic, and mean arterial pressure (MAP). Square wave tests were performed by intermittently pulling the fast flush tab of the pressure transducer to ensure that the dynamic response of the system was adequate and 2 wave deflections after the square wave on the screen of the monitor were present. The thermistor located at the tip of the catheter was connected to the same monitor for recording ΔT_blood during TPTD_CO measurements.

A 7‐french, 110 cm long, double lumen, thermistor‐tipped catheter⁸ was inserted into a jugular vein through an 8.5 F introducer sheath and advanced until its tip was positioned in the pulmonary artery using pressure waveform guidance. The proximal and distal ports of the catheter were connected to fluid‐filled pressure transducers⁹ zeroed at the level of the heart for continuous display of central venous pressure (CVP) and pulmonary artery pressure. The thermistor located at the tip of the pulmonary artery catheter was connected to the monitor¹⁰ for recording ΔT_blood during PATD_CO measurements.

Experimental Procedure—Phase‐1

This study was divided in 2 phases (Fig 1A and B). Phase‐1 aimed to evaluate the effect of 2 different volumes of thermal indicator on the agreement and trending ability between TPTD_CO and PATD_CO. Syringes prefilled with 5 and 10 mL of physiological saline were immersed in an iced bath for at least 1 hour before commencing CO measurements. For each data sampling time, CO was averaged from 3 repeated measurements using 5 mL and 10 mL of ice‐cold (2–5°C) physiological saline. The initial order of thermal indicator volumes was determined at random; this order was inverted during every other subsequent measurement. Each thermal indicator bolus was rapidly injected into the CVP port of the pulmonary artery catheter (<3 seconds) for recording of ΔT_blood in the pulmonary artery and in the femoral artery for PATD_CO and TPTD_CO measurements, respectively. Temperature of the injectate was monitored by 2 in‐line thermistors connected to the PATD_CO and TPTD_CO monitors. Thermistors were placed in series between the syringe containing the thermal indicator and the CVP port of the pulmonary artery catheter. Computation constants for CO measurements were based on the catheter type and volume/temperature of the thermal indicator, as recommended by the manufacturers.

Figure 1.

Experimental protocol during Phase‐1 (A) and Phase‐2 (B). PATD_CO, pulmonary artery thermodilution; TPTD_CO, transpulmonary thermodilution cardiac output; PCA_CO, calibrated pulse contour analysis cardiac output.

Low normal CO states were recorded during anesthesia with isoflurane alone (1.5 MAC), and isoflurane combined with 2 ascending constant rate infusions (CRIs) of remifentanil¹¹ (0.3 and 0.6 μg/kg/min). High CO states were induced by 2 ascending dobutamine¹² CRIs (2.5 and 5.0 μg/kg/min) administered during isoflurane anesthesia. The sequence of drug administration was determined at random. Cardiac output was recorded during anesthesia with isoflurane alone and after each CRI was maintained for at least 15 minutes. After data collection, infusion drugs were stopped and a 15‐minute washout period was allowed before commencing the next CRI.

Experimental Procedure—Phase‐2

After the end of Phase‐1, infusion drugs were interrupted and anesthesia was maintained with 1.5 MAC of isoflurane alone for 30 minutes before commencing Phase‐2, which aimed to evaluate the effects of changes in SVR induced by phenylephrine¹³ and nitroprusside¹⁴ on the agreement and trending ability between PCA_CO and PATD_CO.

The PCA_CO was calibrated one single time with 1 TPTD_CO measurement using 10 mL of ice‐cold physiological saline before conditions of increased and decreased SVR were induced by phenylephrine (1.0 μg/kg/min) and nitroprusside (1.0 μg/kg/min), respectively. Further TPTD_CO measurements were not performed during phase‐2 to avoid recalibration of the PCA_CO system. The order of CRI drugs was determined at random and PATD_CO was recorded as the average of 3 repeated measurements using 10 mL of ice‐cold (2–5°C) physiological saline before CRI drugs were administered (baseline) and after each CRI was maintained for 15 minutes. Cardiac output obtained by calibrated pulse contour analysis was recorded simultaneously with each PATD_CO measurement and averaged for comparison. After CO data were obtained, a 15‐minute washout period was allowed before starting the next CRI.

Heart rate (HR) (recorded from a lead II electrocardiogram), MAP (recorded from the femoral artery catheter), PATD_CO, and SVR [(MAP‐CVP)/PATD_CO × 79.9], were recorded at the same times of CO comparisons.

For both phases, CO determinations were accepted only at steady state hemodynamic conditions, defined as HR and MAP recorded immediately before and after each series of CO measurements varying ≤10%.

After the end of data collection, all catheters were removed and a single dose of meloxicam¹⁵ (0.2 mg/kg, IV) was administered for analgesia. The sites of femoral artery and pulmonary artery catheter insertion were manually compressed with ice cubes packed in lap sponges for at least 15 minutes after catheter removal.

Statistical Methods

Commercially available software was used for data analysis and graphic generation,^{16 , 17 , 18 , 19} Cardiac output data recorded during Phase‐1 were analyzed by the Bland Altman method for multiple measurements per subject.¹⁶ Data recorded during Phase‐2 were analyzed separately during baseline (prior to drug infusion), phenylephrine, and nitroprusside infusion by the standard Bland Altman method.¹⁶ Shapiro‐Wilk normality tests were applied to the bias recorded in each phase. During Phase‐1, the bias (difference between TPTD_CO and PATD_CO) recorded during injection of 5 and 10 mL of thermal signal was compared by a Wilcoxon matched‐pairs signed rank test (asymmetrical data distribution).¹⁷ During Phase‐2, the bias between PCA_CO and PATD_CO recorded during baseline, phenylephrine, and nitroprusside CRI were compared by a one‐way repeated measures ANOVA, followed by a Tukey's post hoc test (symmetrical data distribution).¹⁷

During Phase 1 and 2, the percentage bias (PB) was calculated from the mean bias divided by the mean CO (mean of CO values plotted on the X axis of the Bland Altman graphs). The percentage error (PE) was calculated as 1.96 times the SD of the mean bias divided by the same denominator.27 Additionally, the precision of method (POM) of each series of triplicate CO measurements was calculated as 2 times the coefficient of error (CE).28, 29 The CE represents the coefficient of variation divided by the square root of number of replicates (n) (CE = coefficient of variation/√n).28

During Phase‐2, hemodynamic variables recorded at baseline, and during phenylephrine/nitroprusside infusion were compared by an ANOVA for repeated measurements followed by a Tukey's test (P < .05).¹⁷ The arterial blood pressure tracing was inspected for the presence of the dicrotic notch during PCA_CO measurements.

Ability of the test methods to track changes in the reference method was evaluated by 4‐quadrant plot and polar plot analysis.30, 31, 32 Sequential changes in TPTD_CO (ΔTPTD_CO) or in PCA_CO (ΔPCA_CO) were plotted on the y axis, whereas corresponding changes in the reference method (ΔPATD_CO) were plotted on x axis.¹⁷ After dividing the graphs in 4‐quadrants by the intersection of lines originated from zero in both axes, the concordance rate (%) was calculated as the percentage of data in the upper right/lower left quadrants (data that follow the same trend) in relation to the total number of data points.

Sequential changes in CO (ΔCO) were calculated as arithmetic mean of ΔCO of the test and reference method (ΔTPTD_CO + ΔPATD_CO/2 during Phase 1, and ΔPCA_CO + ΔPATD_CO/2 during Phase 2). This mean ΔCO is represented as the distance of the vector from the center of the polar plot.31, 32 The ΔCO values were converted to polar coordinates using a spreadsheet.^{18 ,} 31 The polar angle was calculated as the angle of divergence of the ΔCO from the line of identity.31, 32 The actual polar plots were generated using a graph drawing software.¹⁹ As recommended by Critchley et al.,32 when the mean polar angle was ≤ ±5°, the radial LOA were estimated from the SD of the mean polar angle (±1.96 SD). When the mean angular bias was > ±5°, the radial LOA were estimated from the plot of the inclusion rate against the radial sector size.32

Good, acceptable (or marginal), and poor trending ability was considered if the concordance rates obtained after excluding ΔCO values ≤0.5 L/min were >95%, between 90 and 95%, and <90%, respectivelly.31, 32 Good trending ability based on polar plot analysis was defined by the mean angular bias ≤ ±5° with radial LOA ≤ ±30° obtained after excluding mean ΔCO values ≤0.5 L/min because small changes are likely because of random effects or noise.32

Results

All animals recovered uneventfully from anesthesia. No catheter‐related complications (hematoma, bruising) were observed during the postanesthetic period.

Phase‐1

The time elapsed from placing the catheter introducer until the PATD_CO catheter was positioned in the pulmonary artery was 33 ± 25 minutes. The time for placing the TPTD_CO catheter in the femoral artery was 8 ± 3 minutes. A total of 39 sets of 3 measurements were performed for TPTD_CO and PATD_CO. Minimum and maximum PATD_CO values (mean ± SD) recorded for 10 mL of thermal indicator during phase‐1 ranged from 1.74 ± 0.21 L/min to 6.73 ± 0.70 L/min.

The mean bias and LOA (±1.96 SD) between TPTD_CO and PATD_CO were 0.62 (−0.11 to 1.35) L/min and 0.33 (−0.25 to 0.91) L/min for 5 and 10 mL of thermal indicator, respectively (Fig 2A and B). The bias recorded with 10 mL of thermal indicator was significantly smaller than with 5 mL of thermal indicator (P < .0001). The PB and PE values were 16% and 19% for 5 mL of thermal indicator, respectively. The use of 10 mL of injectate slightly decreased the PB to 9% and the PE to 16%.

Figure 2.

Bland Altman plots for repeated measurements of differences between transpulmonary thermodilution cardiac output (TPTD_CO) and pulmonary artery cardiac output (PATD_CO) using 5 mL (A) and 10 mL (B) of ice‐cold physiological saline as thermal indicator (Phase‐1); and Bland Altman plots of differences between pulse contour analysis cardiac output (PCA_CO) and PATD_CO during phenylephrine‐induced vasoconstriction (C) and nitroprusside‐induced vasodilation (D) (Phase‐2). Each dog is represented by different geometric figures. Mean biases and limits of agreement (±1.96 SD) are shown by continuous and dashes lines, respectively.

The POM of PATD_CO was 4.2% and 3.1% and the POM of TPTD_CO was 3.7% and 3.4% for 5 and 10 mL of ice‐cold injectate, respectively.

Analysis of trending ability during Phase‐1 is presented in Table 1 and Figure 3A–D. Concordance rates were >95% regardless of the thermal indicator volume. The mean angular bias and radial LOA (calculated as ±1.96 SD) were < ±5° and < ±30°, respectively, for both volumes of ice‐cold injectate.

Table 1.

Variables derived from 4‐quadrant and polar plot analysis to assess the ability of transpulmonary thermodilution (TPTD_CO) and of calibrated pulse contour analysis (PCA_CO) for estimating changes in cardiac output measured by pulmonary artery thermodilution (PATD_CO). Trending analysis was performed using 5 and 10 mL of ice‐cold physiological saline as thermal indicator during simultaneous measurements of TPTD_CO and PATD_CO. Trending analysis between PCA_CO and PATD_CO was performed using 10 mL of ice‐cold physiological saline solution for the thermodilution method. Only data after excluding mean cardiac output changes ≤0.5 L/min are presented

	Phase‐1: TPTDCO and PATDCO (5 mL of Thermal Indicator)	Phase‐1: TPTDCO and PATDCO (10 mL of Thermal Indicator)	Phase‐2: PCACO and PATDCO
Four‐quadrant plot analysis
Measurements in the right quadrant (n)	23	23	10
Measurements in the wrong quadrant (n)	0	0	6
Concordance rate (%)	100	100	63
Polar plot analysis
Measurements with angular bias ≤30° (n)	23	23	3
Measurements with angular bias >30° (n)	0	0	8
Mean angular bias (°)	2	−1	38
Radial limits of agreement (°)	−10 to 14	−9 to 6	5–71

Figure 3.

Four‐quadrant and polar plots of changes in transpulmonary thermodilution cardiac output (ΔTPTD_CO) in relation to changes in pulmonary artery thermodilution (ΔPATD_CO) using 5 mL (A and B) and 10 mL (C and D) of ice‐cold physiological saline as thermal indicator (Phase‐1); and 4‐quadrant and polar plots of changes in pulse contour analysis cardiac output (ΔPCA_CO) in relation to ΔPATD_CO (Phase‐2) during phenylephrine and nitroprusside‐induced changes in systemic vascular resistance (E and F). The exclusion zones (cardiac output changes ≤0.5 L/min) in 4‐quadrant plots (A, C, and E) are shown by the intersection of the dashed lines whereas in the polar plots they are shown by the dashed circles. Mean angular biases and radial limits of agreement of polar plots are shown by continuous and dashed lines, respectively (B, D, and F).

Phase 2

A total of 24 sets of 3 measurements were performed for PCA_CO and PATD_CO comparisons. Compared with baseline, phenylephrine significantly (P < .05) increased SVR and MAP by 44% and 18%, respectively, whereas PATD_CO, PCA_CO and HR were significantly decreased by 18%, 20% and 10%, respectively. Nitroprusside significantly decreased SVR, MAP and PATD_CO by 14%, 27% and 15% from baseline, respectively, without significant changes in HR and PCA_CO (Table 2).

Table 2.

Mean ± SD values of heart rate (HR), pulmonary artery thermodilution cardiac output (PATD_CO), pulse contour analysis cardiac output (PCA_CO), mean arterial pressure (MAP), and systemic vascular resistance (SVR) recorded in 8 isoflurane anesthetized dogs during phase 2 of the study, before drug administration (Baseline), during phenylephrine (1 μg/kg/min), and nitroprusside administration (1 μg/kg/min)

Variable	Baseline	Phenylephrine	Nitroprusside
HR (beats/min)	131 ± 12a	118 ± 8b	130 ± 8a
PATDCO (L/min)	5.2 ± 0.8a	4.2 ± 0.6b	4.4 ± 1.0b
PCACO (L/min)	4.9 ± 0.8a	3.9 ± 0.8b	6.0 ± 1.4a
MAP (mmHg)	88 ± 14a	104 ± 9b	64 ± 15c
SVR (dynes/s/cm5)	1325 ± 74a	1906 ± 190b	1134 ± 153c

^a,b,cMeans followed by different superscript letters are significantly different from each other (Tukey's, P < .05).

The dicrotic notch was clearly identified during phenylephrine CRI in all animals. During nitroprusside CRI, the incisura of the dicrotic notch was unidentifiable at most instances (Fig 4).

Figure 4.

Arterial blood pressure tracings (25 mm/s) obtained from a dog anesthetized with isoflurane during intravenous phenylephrine (1 μg/kg/min) (A) and nitroprusside (1 μg/kg/min) infusions (B). The incisura of the dicrotic notches is shown by arrows.

Mean bias (LOA) between PCA_CO and PATD_CO at baseline were 0.47 (0.18–0.76) L/min, while the PB and PE values at this time were 11% and 6%, respectively. Mean bias (LOA) between PCA_CO and PATD_CO were 0.22 (−0.63 to 1.07) L/min during phenylephrine CRI and 2.12 (0.70–3.55) L/min during nitroprusside infusion (Fig 2C and D). The bias recorded during phenylephrine CRI did not differ from the bias at baseline (P = .15); whereas the bias recorded during nitroprusside CRI was significantly higher in comparison to the bias recorded at baseline (P = .0005) and to the bias recorded during phenylephrine infusion (P = .0006). The PB and PE recorded at baseline was 10% and 6%, respectively. Phenylephrine CRI decreased PB to 6% and increased the PE to 23%; whereas nitroprusside CRI increased the PB to 43% and the PE to 29%. The POM of PATD_CO was 3.5% and POM of PCA_CO was 2.6%.

During Phase‐2, the concordance rate was <90% (Table 1 and Fig 3E). The mean angular bias was > ±5° and the radial LOA was > ±30° (Table 1 and Fig 3F).

Discussion

The main findings of the present study were: (1) The TPTD_CO technique can be used in dogs in replacement of PATD_CO because it provides good agreement and trending ability over a wide range of CO values. (2) The PCA_CO has limited clinical application because it does not agree well with PATD_CO and has a poor ability to track changes in PATD_CO during acute changes in SVR. The slightly better agreement and trending ability observed with 10 mL in comparison to 5 mL of thermal indicator corroborates the recommendation for the use of larger volumes of thermal indicator to improve the reliability of TPTD_CO measurements.2, 5 The trend for larger overestimation of PATD_CO by TPTD_CO with the lower volume of ice‐cold indicator can be explained by greater loss of cold indicator by conductive warming as the solution traverses longer distances between the injection point (vena cava/right atrium) and the site of ΔT_blood measurement (pulmonary artery for PATD_CO versus femoral artery for TPTD_CO).2, 5, 6 Because CO is inversely proportional to the area under the thermodilution curve according to the Stewart‐Hamilton equation, an artifactual decrease in the area under the thermodilution curve because of conductive warming will lead to an overestimation of CO values.2, 5, 6 The worsening of the agreement between PCA_CO and PATD_CO during vasodilation and the poor trending ability between these methods during acute changes in SVR can be explained by the influence of changes in the vasomotor tone on the shape of the arterial pressure wave, which might not be recognized and properly corrected by the PCA_CO algorhythm.18

In the present study, the use of 10 mL of ice‐cold indicator for TPTD_CO measurements resulted in smaller LOA (−0.25 to 0.91 L/min) and in a smaller PE (16%) than the LOA (−0.86 to 0.78 L/min) and an estimated PE of 36% in an earlier study in dogs weighting 12–18 kg using the same volume of ice‐cold indicator.16 In another recent study, PATD_CO and TPTD_CO measurements with 10 mL ice‐cold injectate in beagle dogs resulted in even wider LOA (−2.37 to 2.29 L/min/m²).17 The authors incorrectly concluded that TPTD_CO agreed well with PATD_CO in spite of the wide LOA.17 Percentage error estimated from the data presented in that study17 was >50%, likely representing a poor precision of agreement between the test and reference methods (POA_TESTxREF). The PE, calculated from the Bland Altman plots, is superior to the LOA for estimating the POA_TESTxREF because its calculation incorporates an estimation of the “true” CO values from the Bland Altman plots (mean CO).27, 29 Although the use of PE values to compare CO studies has been recommended since 1990s,27 several publications evaluating the agreement between TPTD_CO and PCA_CO with PATD_CO published thereafter have not reported the PE.8, 9, 10, 11, 12, 14, 16, 17, 18, 19, 23, 24

Although some studies have defined that the acceptable PE of the new method should be ≤30% if it is to be used clinically in replacement for the reference standard,13, 15, 25 this criterion was not adopted in the present study. The use of a predefined PE as a benchmark to determine the interchangeability between the test and the reference methods of CO measurement assumes that the precision of the test method (POM_TEST) should be at least equal to the precision of the reference method (POM_REF). Based on this assumption, the acceptable PE would represent an estimation of the precision of agreement expected if the reference technique was compared to itself (POA_REFxREF), according to the formula: $\text{PE}=\sqrt{{({\text{POM}}_{\mathrm{REF}})}^2+{({\text{POM}}_{\mathrm{REF}})}^2}$.28, 29 Considering the acceptable PE of ≤30% as a benchmark to determine the interchangeability between methods of CO measurement, TPTD_CO and PCA_CO would be considered interchangeable with PATD_CO in the present report. However, the 30% limit is estimated from the precision of single PATD_CO measurements as the reference method (20%)27, 28, 29 and was incorrectly used in studies where PATD_CO was averaged from triplicate measurements.13, 15, 25 The error of PATD_CO can be substantially smaller if values are averaged from repeated measurements and if there is a strict control of factors that increase the variability of repeated PATD_CO measurements (absence of arrhythmias, cardiovascular stability, constant injectate temperature, and volume).

More recently, it has been recommended to determine the acceptable PE based on the POM_REF calculated for each study (POA_REFxREF), as described above, and to compare these values with the PE calculated from the Bland Altman plots, which represents the POA_TESTxREF.28, 29 If the POA_TESTxREF was worse than the POA_REFxREF, then the POM_TEST would have been worse than the POM_REF and the interchangeability between the new method and the reference standard would have been rejected.28, 29 However, if the acceptable PE was estimated from the POM_REF obtained in our study as it has been recommended,28 the acceptable PE would range from 4 to 6% and the interchangeability between the test method of Phase‐1 (TPTD_CO) and the reference standard (PATD_CO) would have been paradoxically rejected because the PE derived from the Bland Altman plots was larger (19 and 16% for 5 and 10 mL of thermal indicator, respectively). However, this apparent discrepancy can be explained by the fact that the PE obtained from Bland Altman plots (POA_TESTxREF) is composed not only by the combination of the POM_TEST and the POM_REF, but also by the general variability of CO measurements about the true CO values (defined as trueness).29 For the reasons explained above, we determined the POM_TEST and POM_REF from repeated measurements and did not use the PE estimated from the POM_REF to accept or reject the tested methods because calculation of the acceptable PE from the POM_REF leads to falsified conclusions about the POM_TEST as it does not take into account the method's variability about the true values (trueness).29

The use of 10 mL, rather than 5 mL of thermal injectate, improved the accuracy of agreement (AOA) and the POA_TESTxREF (smaller PB and PE, respectively) but did not result in obvious differences in POM values. Calculation of the POM requires that the true value is held constant over time during repeated measurements. Considering that an exactly constant true value is impossible to be achieved to allow POM calculation in live animal models,29 we assumed in our study design that the true CO value was relatively constant during each series of triplicate CO measurements because CO values were recorded under conditions of cardiovascular stability (defined as HR and MAP recorded at the beginning and at the end of each triplicate CO measurements differing by ≤10%). The relatively good POM recorded during triplicate CO measurements throughout the study (error <5%) shows that the overall precision of all methods was good and suggests that true CO values were held relatively constant to allow the recording of repeated measurements.

Although both volumes of thermal indicator resulted in good overall agreement and good trending ability between TPTD_CO and PATD_CO, the use of 10 mL of ice‐cold indicator might be preferred in some instances because it significantly decreased the bias and provided a slightly better POA_TESTxREF (lower PE) and a slightly lower tendency for TPTD_CO to overestimate PATD_CO (PB closer to zero). On the other hand, the use of 5 mL boluses of thermal indicator might be favored if there are concerns with fluid overload or if several TPTD_CO measurements are required during short periods of time.

The injectate volume ranged from 0.16 to 0.24 mL/kg and from 0.32 to 0.48 mL/kg for 5 and 10 mL of ice‐cold injectate, respectively. These volumes were above the minimum volume recommended for accurate PATD_CO measurements in human infants (0.15 mL/kg).33 To minimize the greater loss of thermal signal with the transpulmonary technique, the injectate volumes used for TPTD_CO are larger than the volumes recommended for PATD_CO in individuals with the same body mass. The volume of ice‐cold injectate used for TPTD_CO measurements in human pediatric patients in one report was 1.5 mL plus 0.15 mL/kg.34 Based on the volume used for TPTD_CO in human infants,34 3 animals of the present report would have received a slightly larger volume (5.3–6.2 mL of ice‐cold physiological saline) than the fixed 5 mL volume of thermal injectate administered during Phase‐1.

Compared to baseline conditions, phenylephrine administration slightly improved the AOA (PB decreased from 11% to 6%). In spite of a good POM_TEST (2.6%) and POM_REF (3.5%), the POA_REFxTEST was worsened by vasoconstriction (PE increased from 6% at baseline to 23% during phenylephrine). This observation might be at least in part explained by an increase in test method's variability (PCA_CO) around true CO values (trueness).29

During nitroprusside‐induced vasodilatory states, Bland Altman analysis showed an unacceptable overestimation of PATD_CO by PCA_CO. Compared to baseline, the vasodilatory state induced by nitroprusside significantly increased the mean bias from 0.47 L/min to 2.12 L/min and widened the LOA from ± 0.29 L/min to ± 1.43 L/min. Although the AOA was markedly worsened by nitroprusside (PB: 43%), the POA_TESTxREF did not seem to be markedly affected when the vasodilatory state (PE: 29%) was compared to the vasoconstrictive state (PE: 23%). However, the PE values recorded during nitroprusside CRI were underestimated because of the poor AOA between methods. Because PCA_CO values were substantially higher than PATD_CO measurements (mean bias 2.12 L/min), there was an artifactual increase in the denominator of the PE formula (mean CO), leading to an underestimation of PE calculated during the vasodilatory state. Therefore, the PE recorded during nitroprusside CRI would have been even larger if the AOA was not markedly deviated from zero.

Because the agreement between methods was not the same as the hemodynamic state changed from high to low SVR, the ability of PCA_CO to track changes in PATD_CO was also poor as shown by the 4‐quadrant plot and polar plot analysis. These observations, confirm previous reports showing that calibrated pulse contour analysis does not provide reliable agreement with PATD_CO in individuals that present hemodynamic instability induced by vasodilatory states.25

Analysis of the arterial pressure wave showed that the dicrotic notch was not identifiable during nitroprusside‐induced vasodilation in the present study. Vasodilatory states attenuate the rebound of arterial blood (wave reflection) against the closed aortic valve at the end of the ejection phase and this phenomenon could have blunted or attenuated the dicrotic notch during nitroprusside infusion.35 Failure of the algorithm to detect the end of the ejection phase (dicrotic notch) during nitroprusside‐induced vasodilation could lead to an artifactual increase in the area under the systolic portion of the arterial pressure wave, causing overestimation of CO values.18

A limitation of the TPTD_CO method is the requirement for catheterization of a central artery. More peripheral arteries (eg, dorsal metatarsal artery in dogs) would be more practical and carry lower risk of complications, but in one report the TPTD_CO monitor failed to detect a 20 mL bolus of ice‐cold physiological saline when a 22 cm long thermodilution catheter was placed in the dorsal metatarsal artery in 8 out of 30 attempts for calibrating the PCA_CO with TPTD_CO.36 Although no catheter‐related complications were observed in the present report, femoral artery catheterization might result in increased risk of bleeding/hematoma and these potential morbidities should be considered if the technique is used in veterinary patients.

Among the study limitations, one should consider the relatively small number of animals evaluated. However, even with this limited number of animals, clear differences in the agreement and trending ability of the 2 tested methods (TPTD_CO and PCA_CO) against the reference technique (PATD_CO) were evident. Also, this study had some limitations in the design as comparisons between TPTD_CO and PATD_CO were not performed during changes in SVR; whereas comparisons between PCA_CO and PATD_CO were not performed over a wide range of CO values. Because hemodynamic conditions induced during Phase‐1 and Phase‐2 differed substantially, results obtained during both phases might not be directly comparable.

Conclusion

The use of a larger volume of thermal indicator (10 mL versus 5 mL of ice‐cold physiological saline) minimizes the trend for TPTD_CO to overestimate PATD_CO. Although the use of 10 mL of thermal indicator significantly decreases the bias and slightly improves the agreement (AOA and POA) and the trending ability between TPTD_CO and PATD_CO when compared with a lower volume (5 mL) of ice‐cold indicator, both volumes can be used for TPTD_CO measurements in replacement of PATD_CO in dogs.

While phenylephrine‐induced vasoconstriction caused a dual effect on the agreement between calibrated PCA_CO and PATD_CO (increased AOA and decreased POA) without significantly changing the bias, nitroprusside‐induced vasodilation worsened the overall agreement between these methods (significantly larger bias, decreased AOA and POA) and caused a significant overestimation of PATD_CO by PCA_CO. Because the agreement and the ability of PCA_CO to track changes in PATD_CO are poor during drug‐induced changes in SVR, this method is of limited clinical usefulness.