The prognostic role of echo-phenotyping in cardiogenic shock

Background Cardiogenic shock has a variable mortality according to the severity of presentation (SCAI) and the phenotype (SHARC). Pulmonary artery catheter is considered the gold standard for the definition of those. However, the rate of use at the first medical contact is limited. We aim to study the accuracy of CS phenotype definition with echocardiography. Methods We analyzed data from The multicentric, prospective, observational Altshock-2 registry (NCT04295252) from March 2020 to November 2024. CS was defined and stratified according to the SCAI updated criteria. To assess the association between echocardiographic parameters and in-hospital mortality, univariable and multivariable (stepwise) Cox regression analyses were performed, with hazard ratios (HR) and 95% confidence intervals (CI) reported. Spline function for the echocardiographic parameters were made and then Partial Dependence Plot (PDP) or Marginal Effects Plot. Results Among 1056 patients included in the registry, LVEF was available in 784 (74.2%), TAPSE in 641 (60.7%), quantitative TR in 762 (72.2%), and MR in 796 (75.4%). Pre-existing LV dysfunction was present in 23.3% of patients (85% HF-CS, 12.9% ischemic). For this analysis, only patients with complete echocardiographic data (n = 641) were included. The population was divided according to the phenotype as follows: LV predominant (LVEF < 40%); RV predominant (TAPSE ≤ 16 mm); biventricular. Overall in-hospital mortality was 33.5%. Mortality varied by CS phenotype, with 21.5% for isolated LV failure, 44% for biventricular failure, and 57.1% for isolated RV failure (p < 0.001). When stratified by etiology, the interaction term LVEF × etiology was statistically significant (p = 0.037), with LVEF not being associated with in-hospital mortality in non-ischemic patients (p = 0.12), but predictive in ischemic patients (HR 0.980 per 1% increase, 95% CI 0.964–0.997, p = 0.033). Additionally, the end-diastolic diameter was associated with mortality in HF-CS, while it was not in the ischemic subgroup. TAPSE was significantly associated with mortality in both etiologies, with a stronger effect in ischemic patients (ischemic HR 0.897, 95% CI 0.860–0.935; non-ischemic HR 0.944, 95% CI 0.895–0.996), confirmed both at univariate and multivariate analysis. The addition of TAPSE improved the model fit (−2LL decreased to 2400, Δχ² = 29.2, df = 3, p < 0.001) and increased the AUC to 0.78 (95% CI 0.743–0.817), with the DeLong test confirming statistical significance (difference in area = 0.175, SD = 0.0275, z = −6.35, p < 0.0001). By etiology, the combined SCAI B–C + TAPSE model showed AUCs of 0.820 (95% CI 0.775–0.864, p < 0.001) for ischemic CS patients and 0.75 (95% CI 0.698–0.817, p < 0.001) for HF-CS patients, and both were significantly improved compared with SCAI alone (DeLong p < 0.0001) Conclusion RV function is a modifier and its evaluation, either on both invasively and non invasively, should be integrated in the SCAI classification. Univariate e Multivariate analisys