Comparison of Invasive Versus Non-Invasive Pulse Contour-Based Cardiac Output Measurements at Rest and During...

Background/Objectives: Measuring cardiac output (CO) is essential for diagnosis and therapeutic monitoring in pulmonary hypertension (PH). CO assessment based on thermodilution (TD) or Direct Fick (DF) during standard right heart catheterization (RHC) is impractical for regular follow-up. We evaluated the accuracy and agreement of non-invasive Modelflow (MF)-based CO assessment compared with TD and DF during rest and exercise RHC in PH. Methods: This post hoc analysis from a crossover RCT included 24 PH patients (7 females, 59 ± 14 years; mean pulmonary artery pressure 37 ± 11 mmHg) who underwent RHC with repetitive CO assessments at rest and during exercise. CO was measured by TD, DF, and non-invasive MF by fingertip pulse contour analysis at rest and during stepwise cycling to maximal exertion. Results: At rest, mean CO was comparable between methods: TD = 6.05 ± 1.80 L/min, DF = 5.68 ± 1.88 L/min, MF = 6.09 ± 1.84 L/min. At end-exercise, CO increased to TD = 11.18 ± 4.38 L/min, DF = 11.84 ± 4.74 L/min, MF = 8.38 ± 2.93 L/min. Bland-Altman showed minimal bias at rest (MF vs. TD: 0.04 L/min; MF vs. DF: -0.07 L/min) but substantial variability during exercise, with underestimation of CO by MF with increasing workloads (MF vs. TD bias = -2.80 L/min; MF vs. DF bias = -4.38 L/min). Limits of agreement were wide across all workloads. Linear regression confirmed an increasing CO with workload, but MF slope was shallower than TD/DF, suggesting proportional bias. Taffé analysis identified a significant differential (5.847) and proportional bias (0.195) indicative of CO overestimation by MF at low CO and underestimation at high CO. Conclusions: MF group-level agreement is acceptable, but individual-level accuracy is limited, indicating that MF may be suitable for trend monitoring but its applicability for clinical decision-making is restricted, especially during exercise.