Does the method matter? Direct fick vs thermodilution in the haemodynamic assessment of pulmonary hypertension

Dear editor,

We thank Melin et al.¹ for their important work. Collecting more than 852 simultaneous thermodilution (TD) and direct Fick (DF) cardiac output (CO) measurements represents a substantial effort and provides important information for the field. Because DF is technically demanding and time-consuming, TD has largely replaced DF in clinical practice.² However, several studies have revealed the imprecision of TD compared with DF.³ More generally, CO measurements, regardless of the method used, whether invasive or non-invasive show substantial imprecision reflected by wide limits of agreement (LoA) in the range of ±2 to 3 L/min across most measurement methods.³ Appropriately, Melin et al.¹ questioned the potential impact of using DF instead of TD to measure CO on pulmonary hypertension (PH) classification. We similarly estimated this error using mathematical modelling based on the imprecision reported in the literature between DF and TD and observed results comparable with a diagnostic error of 6% across 1142 consecutive RHC,⁴ which corresponds closely to the 94% diagnostic accuracy reported by Melin et al.¹ We believe that several aspects of the results warrant further discussion.

First, Melin et al.¹ reported a 15% over diagnosis of a precapillary component defined by a pulmonary vascular resistance (PVR) >2 WU, when using TD. This observation should be interpreted in the light of the low mean PVR of 2.5 WU reported in their cohort. The risk of misclassification intrinsically depends on the absolute level of PVR and is highest in patients whose measurements lie close to the diagnostic threshold.⁴ These patients typically belong to group 2 or 3 PH. Such misclassification is unlikely to have major clinical consequences, as these patients are most often managed similarly regardless of mild or moderate PH, with treatment primarily targeting the underlying cardiac or pulmonary disease rather than pulmonary arterial hypertension (PAH)-approved drugs.² In contrast, the risk of misclassification in patients with group 1 PAH, who typically present with substantially higher PVR values, is expected to be very low. This point is critical, because these patients benefit from PAH-specific therapies and a diagnostic misclassification could have significant clinical implications. In this context, we believe that the clinical impact of the results of Melin et al.¹ could be further strengthened by providing an analysis of misclassification rates stratified according to clinical PH groups.

Second, the authors concluded that TD overestimates CO in low-flow state and underestimates CO in normal-flow states. We believe that this assumption relies on relatively limited evidence. It is possible that the observed biases of −0.38 L/min in low CO states and +0.22 L/min in normal CO states may simply reflect random variation; this issue could be assessed if the authors provided the 95% CI of the absolute bias.⁵ In high-output states, the bias appears larger (−1.78 L/min), but this finding should be interpreted cautiously given the small number of measurements and the presence of five extreme outliers showing a difference between TD and DF of −6 L/min, which raises concerns regarding the reliability of either DF or TD in these particular cases.

We think that performing DF, which requires additional time and resources, should be considered in rare situations where a meaningful impact on management is anticipated based on changes in CO or PVR. In the field of PH, such situations may include RHC performed within screening programmes, such as in patients with BMPR2 mutations or systemic sclerosis, as well as selected patients with chronic thromboembolic disease and relatively preserved resting haemodynamic. In these populations, PVR values may be only mildly elevated, and measurement imprecision may meaningfully influence diagnostic classification and management decision-making.

In conclusion, the results of Melin et al.¹ highlight the inherent limitations of CO measurement and their potential impact on haemodynamic classification. These findings underscore that excessive reliance on rigid haemodynamic cut-off values for clinical decision-making is likely inappropriate. Careful clinical assessment and thorough evaluation of risk factors before RHC are probably the most important elements for minimizing diagnostic misclassification. Haemodynamic measurements should therefore always be interpreted within the broader clinical context.

Declarations

Disclosure of Interest

All authors have no conflicts of interest directly related to this publication, which does not involve any therapeutic products. Furthermore, the authors have no affiliations with companies marketing devices related to Direct Fick or thermodilution methods. General DOI of the authors are as follow: L.G. has received travel support for attending meeting from MSD. A.B. has received research grants to his institution from Janssen, MSD, Gossamer, consulting fees from MSD, Gossamer and United Therapeutics, speaking fees from MSD, Gossamer, AstraZeneca, Janssen, Ferrer and AOP Health O.S. has received research grants to his institution from Janssen, MSD, Ferrer, AOP Orphan, consulting fees from AOP Orphan, Ferrer, Gossamer Bio, Inhibikase, Janssen, Liquidia, MSD, Pulmovant, Respira Therapeutics, United Therapeutics, Payment or Honoria for lectures, presentations, speaker bureaus, manuscript writing or educational events from AOP Orphan, Janssen, Ferrer, MSD and participated on a data safety monitoring board or advisory board from Gossamer Bio, Janssen, Respira Therapeutics. In the last 3 years. D.M. has received consulting fees from Ferrer, Merck MSD, Gossamer Bio, Payment or Honoria for lectures, presentations, speaker bureaus, manuscript writing or educational events from AOP health France, Bayer, Boerhinger, Chiesi, Gossamer Bio Inc, GSK, Jazz Pharmaceuticals, Ferrer, Merck MSD, support for attending meetings and/or travel from Merck MSD, Oxyvie.

Data Availability

No data were generated or analysed for this manuscript.

Funding

All authors declare no funding for this contribution.