Cardiac output monitors in septic shock: do they deliver what matters? A systematic review and meta-analysis

To evaluate the interchangeability of cardiac output (CO) monitoring devices compared to reference methods in adult ICU patients with septic shock, we systematically searched electronic databases through January 2025 for prospective studies comparing CO monitors with pulmonary artery catheter (PAC), transpulmonary thermodilution (TPTD), or echocardiography. Eligible studies included Bland-Altman analysis and, when available, trending assessment via polar or 4-quadrant plots, precision, and time response. Agreement was defined as percentage error (PE) < 30%, and acceptable trending as concordance ≥ 90%. Pooled bias, limits of agreement (LoA), and PE were calculated using the Sidik-Jonkman random-effects model. Twenty-six studies were included, yielding 37 unique device-reference datasets and encompassing 1,323 patients. PAC was the most common reference (18 datasets), followed by TPTD (16) and echocardiography (3). The pooled bias was 0.15 L min⁻¹ with LoA of ± 3.45 L min⁻¹ and pooled PE of 49%. Calibrated pulse contour analysis (PCA) showed the best agreement (PE 25%), whereas uncalibrated PCA, thoracic electrical bioimpedance, and bioreactance demonstrated poor agreement (PE ≥ 52%). Heterogeneity for mean bias was high across all subgroups (I² >80%). Of 15 datasets reporting trending, only three achieved concordance ≥ 90%. Most CO monitors demonstrate poor agreement with reference methods in septic shock. However, their true clinical utility remains unclear, as usual validation frameworks-centered on Bland-Altman analysis-overlook metrics that matter most to intensivists. Precision, time response, and trending ability are critical for real-time decision-making but were rarely assessed. Future studies must incorporate these parameters to meaningfully evaluate device performance at the bedside. PROSPERO registration: CRD42024509384.