Cardiac Output During Exercise: Thermodilution Versus Direct Fick

Abstract

Background

Accurate cardiac output (CO) measurements during exercise are essential for the diagnosis of exercise pulmonary hypertension (PH), exercise precapillary PH, and exercise postcapillary PH. The purpose of this study is to compare performance of thermodilution CO to gold‐standard direct Fick CO (dfCO) at rest and exercise.

Methods

A single‐center prospective cohort study of patients undergoing invasive cardiopulmonary exercise test over a 3‐year period. For the primary outcome, we predicted CO at each stage of exercise and recovery using generalized additive modeling. In secondary analysis, we assessed mean differences in CO across exercise stages and Bland–Altman analysis at rest and peak exercise. Finally, we assessed the impact in classification of exercise PH, exercise precapillary PH, and exercise postcapillary PH between the 2 CO methods.

Results

A total of 302 patients were included. In the primary analysis, the generalized additive model smooth term was significantly different between thermodilution and dfCO (P<0.001), with thermodilution underestimating dfCO at rest and overestimating dfCO at max exercise. Wide limits of agreement were noted between thermodilution and dfCO, particularly at peak exercise (−0.08 (−4.5 to +4.35) L/min). The classification of exercise PH, exercise precapillary PH, and exercise postcapillary PH was not significantly different between CO methods (4, 3, and 1 patient misclassified by thermodilution, respectively). CO reserve using thermodilution was overestimated compared with dfCO (96.66% versus 88.45%, P<0.001).

Conclusions

There are differences and wide limits of agreement between thermodilution and dfCO both at rest and during exercise. Sequential measurement of thermodilution during exercise with computing of mPAP/CO and PAWP/CO slopes, reduced the clinical impact of the differences in the diagnosis of exercise PH or exercise postcapillary PH.

Nonstandard Abbreviations and Acronyms

CO

cardiac output

CPET

cardiopulmonary exercise test

df

direct Fick

GAM

generalized additive model

mPAP

mean pulmonary artery pressure

PAWP

pulmonary artery wedge pressure

PH

pulmonary hypertension

Clinical Perspective.

What Is New?

• Classification of exercise pulmonary hypertension using slope equations (mean pulmonary artery pressure divided by cardiac output, for example) was not significantly different using thermodilution cardiac output compared with direct Fick cardiac output in patients undergoing invasive cardiopulmonary exercise testing.

What Are the Clinical Implications?

• In cath laboratories without direct Fick cardiac output capabilities, thermodilution may be adequate to diagnose exercise pulmonary hypertensive diseases if sequential measurements and slopes are used.

• Thermodilution and direct Fick cardiac output have wide limits of agreement at rest and during exercise, so single stage measurements should not be considered interchangeable.

Exercise right heart catheterization is commonly performed to evaluate patients with unexplained dyspnea or exercise intolerance. This test is essential to diagnose exercise pulmonary hypertension (ex‐PH), exercise postcapillary PH (ex‐pcPH), and preload insufficiency. ¹ For all these diagnoses, an accurate cardiac output (CO) measurement is essential both at rest and during exercise, since it allows for the determination of mean pulmonary artery pressure (mPAP)/CO and pulmonary artery wedge pressure (PAWP)/CO slopes as well as the CO reserve. In addition, CO measurement allows for the calculation of cardiac index, stroke volume, and pulmonary vascular resistance (PVR) at rest and during exercise.

The gold standard for CO assessment is the direct Fick CO (dfCO) method, which requires the measurement of oxygen consumption with a metabolic cart and the simultaneous determination of arterial and venous oxygen content, under the same conditions. ² , ³ For the arterial and venous oxygen content, both arterial and mixed venous blood gas need to be analyzed with co‐oximetry with their respective measurement of hemoglobin. Given this complexity, CO is commonly determined using thermodilution, a methodology that is widely available and can be easily used during rest and exercise.

At rest, thermodilution has high accuracy compared with dfCO but wide limits of agreement. ² However, at rest, the thermodilution determination is obtained in triplicate with measurements recorded only if there is <10% variation. This methodologic requirement for thermodilution is not possible during exercise because the time limitations of each exercise stage reduce the ability to repeat thermodilution both in ramp and step protocols. We previously contrasted CO methodologies during exercise in a smaller number of patients and noted that thermodilution compared with dfCO had wide limits of agreement during exercise. ⁴

Whether thermodilution is sufficiently accurate to be used during exercise right heart catheterization remains unknown. Therefore, we tested this question in a large prospective cohort of patients who underwent upright invasive cardiopulmonary exercise testing (iCPET) with simultaneous measurement of thermodilution and dfCO at rest and at every stage of exercise and recovery. Given the inherent limitations of thermodilution and the fact that determinations are usually single given time constrains of exercise stages, we hypothesized that thermodilution would be inaccurate during exercise with wide limits of agreement and these variations may have implications for the diagnosis of ex‐PH, ex‐pcPH, and ex‐precapillary PH, conditions described by slopes or calculations that include CO.

METHODS

Data Availability Statement

The data underlying this article cannot be shared publicly due to the limitations imposed by the institutional review board. The data will be shared on reasonable request to the corresponding author after creating a data use agreement between institutions.

Study Design, Participants, and Setting

We performed a single‐center, prospective study, in consecutive patients referred for iCPET from March 2021 to August 2023. The iCPET adds to a traditional exercise right heart catheterization a radial arterial line and CPET. ⁵ The radial line and the pulmonary artery catheter allow for the simultaneous acquisition of arterial blood gases (ABG) and mixed venous blood gases (VBG), respectively. The CPET provides important measurements including the oxygen consumption at the same time of ABG and VBG acquisition. A subset of these patients (n=24) were included in a previous study that tested the value of a noninvasive methodology, pulse contour analysis, in CO determination during exercise. ⁴ The current study was approved by the institutional review board (#16‐872) and informed consent was obtained from every patient. Findings are reported following the Strengthening the Reporting of Observational Studies in Epidemiology guidelines. ⁶

Exercise Protocol

Patients underwent maximal exercise in a sitting cycle ergometer, pedaling at a cadence of 60 revolutions/min using a step protocol with increments of 20 Watts every 2 minutes. Hemodynamic determinations including right atrial pressure, PAP, and PAWP were obtained at all stages. Pulmonary pressures were averaged across the respiratory cycle to account for pressures swings associated with changes in intrathoracic pressure during exercise.

At baseline, in upright position, thermodilution was performed in triplicate to obtain a CO within 10% of one another. Due to limited duration of the exercise stages (2 minutes) and all the determinations involved (pulmonary pressure and VBG), we obtained only 1 thermodilution measurement during each stage of exercise. The VBG was obtained from the distal port of the pulmonary artery catheter, and after this blood was collected, flushing with normal saline was needed to avoid clotting and pressure waveform damping. This flushing made the temperature sensor unstable and not immediately ready for thermodilution, until the blood temperature sensed returned to a stable state. The thermodilution values were on rare occasions repeated during exercise given the time limitations of 2 minutes per stage; but we used a systematic method to decide when to repeat the thermodilution determination, based on the following rules: (1) CO should increase with exercise, therefore any thermodilution value lower than the prior exercise stage was repeated and (2) the increase in thermodilution is progressive through the exercise, therefore a value >50% of the one obtained on prior stage was repeated. Recovery was assessed at 1 and 3 minutes post exercise.

At baseline and at every stage of exercise and recovery, we recorded thermodilution. Immediately after recording thermodilution, 2 nurses simultaneously obtained ABG and VBG after discarding 3 mL of blood, to account for 3 times the dead space between the tip of the catheter and the collection syringe (Portex, Smiths Medical ASD Inc, MN, USA). Extra care was taken to remove any bubble of air, to prevent falsely elevated oximetric determinations, and to turn/rotate the syringe to prevent clotting. ABG and VBG were immediately sent to the laboratory using a pneumatic tube transport system. Blood was immediately processed using a blood gas analyzer with co‐oximetry (Radiometer America ABL800 Flex+ system, CA, USA). At the same time of obtaining ABG and VBG, we carefully recorded the oxygen consumption (Ultima CPX, MGC Diagnostics, Saint Paul, MN). A detailed explanation of our iCPET protocol was described in a prior publication. ⁷

Statistical Analysis

Continuous variables are presented as median (interquartile range) and categorical variables as counts (percentages) unless otherwise specified. Paired t tests were used to assess mean differences in thermodilution and dfCO. We fitted generalized additive models (GAMs) for both CO methods, to assess the primary aim of this study—the difference between thermodilution and dfCO across all stages of exercise, including baseline and recovery. We fit a GAM for both CO methods across exercise stages, because we expected nonlinear trends in measurements as previously observed. ⁴ , ⁸ We included baseline measurements as well as incremental exercise stages at every 20 Watts up to the patient’s maximal tolerance work load (peak exercise). We also included recovery stages at 1 and 3 minutes post exercise. A random effect was included to adjust for patient‐specific variability. Exercise stages were mapped onto a numerical scale ranging from 0 (baseline) to 1 (maximum stage), with intermediate stages evenly spaced. This scale was used to normalize the variable number of exercise stages across patients. Recovery stages were standardized to 1.25 for 1‐minute recovery and 1.5 for 3‐minute recovery, assuming equal‐spacing between intervals. The GAM model employed does not treat the baseline CO values as the mean of the baseline data points. Instead, it uses a smoothing function to capture the overall relationship between each CO and exercise stage for every patient. Therefore, the GAM prediction is not simply a summary of baseline data but an estimate of how baseline CO fits into the overall relationship between CO and exercise stages. More details on the GAM construction are included in Data S1.

Secondary analyses included mean differences between CO methods at baseline and across the first 4 stages of exercise (which were completed by most patients), as well as the maximum exercise stage. We constructed Bland–Altman plots to assess mean differences and 95% limits of agreement, with dfCO as the gold ‐standard. We also report on the impact of thermodilution determination the diagnosis of ex‐PH, ex‐pcPH, and exercise precapillary PH (ex‐prePH), considering dfCO as the gold standard. We computed slopes for mPAP/CO and PAWP/CO using all exercise stages. A mPAP/CO slope >3 Wood units (WU) was used to diagnose ex‐PH and a slope of PAWP/CO >2 WU was used to diagnose ex‐pcPH. A PVR >2 WU at peak exercise was used to identify ex‐prePH. ⁹ We also contrasted the CO reserve, by using thermodilution or dfCO, with paired t test. Reduced CO reserve is observed in preload insufficiency. ⁵ , ¹⁰ To assess potential confounders on thermodilution accuracy, we collected the heart rhythm (ie, the presence of atrial fibrillation) and whether there was intracardiac shunting reported by echocardiography with agitated saline. A 2‐sided alpha of <0.05 was considered significant. All analyses were performed using R Statistical Software (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Patient Characteristics

A total of 302 patients were enrolled, 202 (67%) female, median age 49 (interquartile range, 37–63) years, with body mass index 30.24 (interquartile range, 25.96–36.59) kg/m² (Table 1). No patients had more than mild tricuspid regurgitation on echocardiography. A summary of hemodynamic data at baseline and across all exercise stages is included in Table 2. There were no missing data across exercise stages. Only 2 patients had atrial fibrillation. Of the 130 patients who had echocardiography with agitated saline, only 25 (19.2%) showed evidence of intracardiac shunt. No patients had step‐ups in respiratory exchange ratio, end‐tidal CO₂ or end‐tidal O₂, to suggest a patent foramen ovale that opened with exercise.

Table 1.

Patient Characteristics

Characteristic	No.=302
Age, y	49 (37, 63)
Female sex (%)	202 (67%)
Height, cm	167.6 (162.6, 175.3)
Weight, kg	87.5 (72.9, 103)
Body surface area, m2	1.99 (1.80, 2.13)
Body mass index, kg/m2	30.24 (25.96, 36.59)
Resting PH, yes	37 (12.3%)
Postprocedure diagnoses (n=265)*
Preload insufficiency	132 (49.8%)
Exercise precapillary PH	43 (16.2%)
Exercise PH	41 (15.5%)
Exercise postcapillary PH	4 (1.5%)

Median (interquartile range) unless otherwise specified.

PH indicates pulmonary hypertension.

^*Patients without resting pulmonary hypertension.

Table 2.

Hemodynamics by Exercise Stage

Characteristics	Baseline	20 Watts	40 Watts	60 Watts	Peak	1‐min recovery	3‐min recovery
Pulse oximetry (%)	98.00 (97.00 to 99.00)	98.00 (96.00 to 98.00)	98.00 (96.00 to 98.00)	98.00 (95.00 to 98.00)	97.00 (94.00 to 98.00)	98.00 (96.00 to 99.00)	98.00 (96.00 to 98.00)
Heart rate, bpm	82.00 (73.00 to 91.75)	94.00 (85.00 to 105.00)	101.00 (92.00 to 116.00)	112.00 (101.00 to 127.00)	148.00 (129.25 to 164.75)	130.00 (115.75 to 147.25)	109.00 (98.25 to 120.75)
Systolic BP, mm Hg	149.00 (138.00 to 165.00)	158.00 (141.25 to 171.00)	167.00 (153.00 to 185.50)	176.00 (162.00 to 199.00)	195.00 (180.00 to 220.00)	158.00 (136.50 to 180.00)	148.00 (131.00 to 167.00)
Diastolic BP, mm Hg	79.00 (72.00 to 87.00)	76.00 (69.00 to 83.75)	78.00 (71.00 to 87.00)	79.00 (71.00 to 88.00)	82.00 (74.00 to 92.00)	72.00 (62.00 to 79.00)	70.00 (61.00 to 77.00)
Right atrial pressure, mm Hg	2.00 (0.00 to 4.00)	3.00 (0.00 to 5.00)	3.00 (0.00 to 5.00)	3.00 (0.00 to 6.00)	2.00 (0.00 to 5.00)	1.00 (0.00 to 3.00)	0.00 (−1.00 to 3.00)
Systolic PAP, mm Hg	19.00 (15.00 to 23.00)	23.00 (19.00 to 29.00)	26.00 (21.00 to 35.00)	28.00 (23.00 to 38.00)	35.50 (29.00 to 45.00)	28.00 (22.00 to 37.00)	22.00 (18.00 to 28.75)
Diastolic PAP, mm Hg	9.00 (6.00 to 12.00)	11.00 (8.00 to 15.00)	11.00 (8.00 to 16.00)	12.00 (9.00 to 17.00)	13.50 (8.00 to 19.00)	10.00 (5.00 to 16.00)	9.00 (5.00 to 13.00)
Mean PAP, mm Hg	13.50 (10.00 to 17.00)	17.00 (13.00 to 22.00)	19.00 (15.00 to 25.00)	21.00 (16.00 to 28.50)	26.00 (20.00 to 31.00)	20.00 (15.00 to 27.00)	15.00 (12.00 to 20.00)
Pulmonary artery wedge pressure, mm Hg (mean)	3.00 (1.00 to 5.00)	5.00 (3.00 to 8.50)	6.00 (4.00 to 10.00)	7.00 (4.00 to 10.00)	8.00 (6.00 to 13.00)	4.00 (2.00 to 7.00)	3.00 (1.00 to 5.00)
Pulmonary vascular resistance, Wood units	2.00 (1.56 to 2.63)	1.62 (1.25 to 2.18)	1.55 (1.16 to 2.35)	1.44 (1.10 to 1.88)	1.16 (0.88 to 1.72)	1.72 (1.24 to 2.46)	1.46 (1.16 to 2.10)
Thermodilution CO, L/min	4.90 (4.10 to 6.00)	6.90 (5.60 to 8.40)	8.00 (6.60 to 9.70)	9.40 (8.00 to 10.80)	13.70 (11.25 to 16.40)	9.00 (7.20 to 11.20)	7.90 (6.40 to 9.70)
dfCO, L/min	5.10 (4.20 to 6.20)	7.30 (6.13 to 8.40)	8.25 (7.03 to 9.48)	9.40 (8.20 to 10.70)	12.60 (10.70 to 15.10)	10.60 (8.70 to 12.80)	8.20 (7.00 to 9.80)
dfCO‐thermodilution, bias, L/min (limits of agreement)	0.1 (−2.23 to +2.42)	0.31 (−2.85 to +3.48)	0.09 (−2.88 to +3.06)	0.02 (−2.97 to +3.01)	−0.08 (−4.5 to +4.35)	1.41 (−3.95 to +6.78)	0.04 (−0.49 to +0.58)

Median (interquartile range), unless specified. The hemodynamic determinations of right atrial pressure, pulmonary artery pressure, pulmonary artery wedge pressure were obtained as averaged values across the respiratory cycle.

BP indicates blood pressure; CO, cardiac output; dfCO, direct Fick cardiac output; and PAP, pulmonary artery pressure.thermodilution.

Hemodynamic Phenotype on iCPET

Resting PH (mPAP >20 mm Hg) was present in 37 (12.3%) patients. Of the remaining 265 patients, 132 (49.8%) were diagnosed with preload insufficiency, 43 (16.2%) with ex‐prePH, 41 (15.5%) ex‐PH, and 4 (1.5%) with ex‐pcPH (Table 1).

Comparison Between Thermodilution and dfCO

We noted significant differences between thermodilution and dfCO with GAMs (P<0.001) (Figure 1). The predicted increase in dfCO per 20 Watts was 0.45 L/min (SD 1.50 L/min). The expected decrease in dfCO from peak to 1‐ and 3‐minute recovery was 1.25 and 3.41 L/min, respectively.

Figure 1. Generalized additive model for comparison of thermodilution and dFCO.

The GAM curves show thermodilution higher at baseline and lower at peak, whereas Table 2 shows the opposite—the GAM curve reflects the model’s adjusted estimates, whereas Table 2 presents unadjusted averages. This difference arises from the model’s smoothing and adjustment for confounders. Overall comparison between methods was assessed by an F‐test of nested GAMs, comparing a model with method‐specific smoothing effect to one with the same smoothing across methods. CO indicates cardiac output; dfCO, direct Fick cardiac output; and GAM, generalized additive model.

Mean differences and limits of agreement for dfCO and thermodilution across exercise stages and recovery are shown in Table 2. The mean difference and 95% limits of agreement between thermodilution and dfCO were +0.1 (−2.23 to +2.42) L/min and −0.08 (−4.50 to +4.35) L/min, at rest and peak exercise, respectively (Figure 2).

Figure 2. Bland–Altman plots comparing thermodilution and dfCO at baseline (A) and peak exercise (B).

We color coded patients with atrial fibrillation with blue dots and patients with intracardiac shunt as yellow dots. The rest of the patients are presented as black dots. AFib indicates atrial fibrillation; dfCO, direct Fick cardiac output; LOA, limits of agreement; NSR, normal sinus rhythm; and tdCO, thermodilution cardiac output.thermodilution

Secondary Outcomes

In the subset of 265 patients with normal resting pulmonary pressures, the mean mPAP/dfCO slope was 1.99 WU compared with mean mPAP/thermodilution slope of 2.18 WU (mean difference, 0.18 [95% CI, −0.25 to −0.13], P<0.001). The mean PAWP/dfCO slope was 0.68 WU compared with mean PAWP/thermodilution slope of 0.71 WU (mean difference, 0.03 [95% CI, −0.05 to 0.002], P=0.07). PVR at max exercise was 1.70 (SD 1.36) WU with thermodilution versus 1.65 (SD 1.26) WU using dfCO and, a nonsignificant difference (mean difference, 0.01 [95% CI, −0.03 to +0.06], P=0.6).

Using dfCO, 41 (15.5%) patients were classified as ex‐PH compared with 45 (17.0%) with thermodilution; 43 (16.2%) patients were classified as ex‐prePH compared with 46 (17.4%) patients with thermodilution, and 4 patients (1.5%) were classified as ex‐pcPH using dfCO compared with 5 (1.9%) with thermodilution (Figure 3). Cardiac output reserve percentage differed between the 2 methods, with thermodilution overestimating reserve compared with dfCO (96.66% versus 88.45%, mean difference, 8.21 [95% CI, 3.77–12.66], P<0.001). The presence of intracardiac shunt did not significantly affect the relationship between CO methods by GAM analysis (P value for the interaction 0.13) (Figure S1).

Figure 3. Classification of exercise phenotypes by thermodilution and dfCO.

Sankey diagram shows the proportion of patients with Ex‐PH, Ex‐prePH, and Ex‐pcPH (blue portions of the bar) when using either dfCO or thermodilution. Misclassification is shown by the number of patients who cross colors between bars. dfCO indicates direct Fick cardiac output; Ex‐pcPH, exercise postcapillary pulmonary hypertension; Ex‐PH, exercise pulmonary hypertension; Ex‐prePH, exercise precapillary pulmonary hypertension; and tdCO, thermodilution cardiac outputthermodilution

DISCUSSION

In the largest study of its kind to date, we demonstrated significant differences between dfCO and thermodilution during exercise using an advanced modeling approach. We noted wide limits of agreement between thermodilution at rest and at peak exercise when compared with the gold standard dfCO. Additionally, there was a statistically significant difference between mPAP/CO slopes when calculated by thermodilution versus dfCO and a borderline difference between PAWP/CO slopes. However, the differences in slopes between CO methods did not meaningfully alter the classification of patients with ex‐PH and ex‐pcPH. The PVR at peak exercise was numerically lower when using dfCO (although not significantly different), and this small variation misclassified a few patients with PVR >2 WU at peak exercise. Finally, CO reserve was higher when calculated using thermodilution compared with dfCO.

This study differs from prior evaluations of thermodilution and dfCO during exercise in the large number of patients prospectively included, the contemporary, rigorous, and sound methodology used, and the inclusion of a cohort in whom the majority had normal resting hemodynamics. Hsu et al ¹¹ reported a larger difference in mPAP/CO slopes (roughly 1 WU higher in the thermodilution group) and maximal exercise CO (2.3 L/min lower in thermodilution), which led to a higher rate of misclassification of patients with ex‐PH. However, their mean mPAP/CO slopes were closer to the discrimination value of 3 WU for ex‐PH than our cohort. In addition, the study by Hsu et al. ¹¹ included 20 patients with normal resting hemodynamics, compared with 265 patients in our study, and used supine bicycle ergometry rather than upright bicycle as in our work. Our group previously demonstrated differences in supine and upright hemodynamics during exercise, such as lower filling pressures and CO, though differences in exercise mPAP/CO and PAWP/CO slopes were not significantly different. ¹² Similar to the Hsu study, we identified wide limits of agreement between thermodilution and dfCO both at rest and during exercise. ⁴ , ¹¹

Although there were numerical differences and wide limits of agreement between thermodilution and dfCO, when performing sequential CO measurements during exercise and using mPAP/CO and PAWP/CO slopes, the methodology for measuring CO had minor clinical implications, a finding that is important to support the use of serial thermodilution determinations during exercise. Given wide limits of agreement, because determinations of thermodilution, that is, peak exercise, are not supported by our data given the potential large margin of error. The use of mPAP/CO and PAWP/CO slopes allows for correction of intrinsic variations in the measurement of thermodilution during exercise and reducing the impact of outliers. A specific value may be higher or lower than the gold standard dfCO; however, when these measurements are repeated there is regression to the mean, minimizing the measurement error. However, when a single thermodilution measurement is taken at a specific point and used in calculations such as baseline and peak exercise PVR, the chances of error are more pronounced with direct clinical implications. Variations in thermodilution compared with dfCO have more pronounced implications when PVR at peak exercise and mPAP/CO and PAWP/CO are close to the discrimination points.

Differences in thermodilution and dfCO during exercise may be related to differences in mixing or circulation of blood during this nonsteady state period or the increased coronary sinus flow with exertion. The most likely source of variability is the inability to perform thermodilution in triplicate due to “short” exercise stages compounded by the refractory period for thermodilution measurement after flushing pulmonary artery catheter ports. In order to increase the accuracy in thermodilution measurements we recommend repeating it 3 times with <10% difference at rest and obtaining thermodilution determination at every stage of exercise. Determination of thermodilution during exercise should be critically examined, and if thermodilution decreases or disproportionally increases from one stage of exercise to the other, then thermodilution should be repeated when possible. To reduce errors thermodilution requires careful determination, including a steady injection of a specific volume (usual 10 mL) at a constant temperature with adequate setup of the measuring system, selecting volume of injection and computation constant (or brand and type of pulmonary artery catheter used if incorporated in the software).

The determination of a reliable dfCO also requires important steps and precautions. These include simultaneous acquisition of ABG and VBG by 2 operators, as oxygenation may change with increasing exercise. Pulse oximetry cannot be used, given large limits of agreement with arterial oxygen saturation. ¹³ Both ABG and VBG need to be done with co‐oximetry to eliminate the impact of carboxy‐ and methemoglobin in the percentage of oxyhemoglobin. ¹³ , ¹⁴ We use the hemoglobin reported on the VBG in each stage, because hemoglobin may also vary during exercise. Oxygen consumption should be measured at the time of BG collection, following established criteria and averaging values over several seconds to reduce inaccuracies in the measurement.

This study has limitations, including its single‐center nature. Although there was no significant misclassification in ex‐PH and ex‐pcPH between CO methods, this study may have been underpowered to detect small differences, given the relatively low prevalence of these conditions in our cohort. The very low number of patients with ex‐pcPH likely reflects the referral base for our iCPET program, which predominantly serves patients with unexplained dyspnea and exercise intolerance, who in great proportion have normal resting pulmonary hemodynamics. We did not specifically record the number of times in which thermodilution was repeated during exercise due to unreliable determinations (eg, drop in CO or unexpected increases), but authors estimated that it was ∼5% of the measurements. Future study in this area could be enriched by including patients with higher pretest probability for 1 of these exercise‐induced conditions. Similarly, extrapolating these findings to a more morbid cohort, or those with significant valvular disease, may not be possible. Notably, the evidence for valve regurgitation causing significant discrepancies in thermodilution versus dfCO is questionable. ¹⁵ We favor dfCO over thermodilution in situations such as valve area calculation at rest, for which the highest accuracy is essential. ¹⁶ Despite these limitations, this study provides support for using thermodilution determination during exercise when using serial measurement and computing mPAP/CO and PAWP/CO slopes.

CONCLUSIONS

There are differences and wide limits of agreement between thermodilution and dfCO both at rest and during exercise. Sequential measurement of thermodilution during exercise with computing of mPAP/CO and PAWP/CO slopes reduced the clinical impact of the differences in the diagnosis of ex‐PH or ex‐pcPH.

Sources of Funding

None.

Disclosures

Adriano R. Tonelli participated in advisory boards of Merck and Janssen. The remaining authors have no disclosures to report.

Supporting information

Data S1. Supplemental Methods

Figure S1