Comparison of risk prediction models in infarct-related cardiogenic shock.

Several prediction models have been developed to allow accurate risk assessment and provide better treatment guidance in patients with infarct-related cardiogenic shock (CS). However, comparative data between these models are still scarce. The objective of the study is to externally validate different risk prediction models in infarct-related CS and compare their predictive value in the early clinical course. The Simplified Acute Physiology Score (SAPS) II Score, the CardShock score, the IABP-SHOCK II score, and the Society for Cardiovascular Angiography and Intervention (SCAI) classification were each externally validated in a total of 1055 patients with infarct-related CS enrolled into the randomized CULPRIT-SHOCK trial or the corresponding registry. The primary outcome was 30-day all-cause mortality. Discriminative power was assessed by comparing the area under the curves (AUC) in case of continuous scores. In direct comparison of the continuous scores in a total of 161 patients, the IABP-SHOCK II score revealed best discrimination [area under the curve (AUC = 0.74)], followed by the CardShock score (AUC = 0.69) and the SAPS II score, giving only moderate discrimination (AUC = 0.63). All of the three scores revealed acceptable calibration by Hosmer-Lemeshow test. The SCAI classification as a categorical predictive model displayed good prognostic assessment for the highest risk group (Stage E) but showed poor discrimination between Stages C and D with respect to short-term-mortality. Based on the present findings, the IABP-SHOCK II score appears to be the most suitable of the examined models for immediate risk prediction in infarct-related CS. Prospective evaluation of the models, further modification, or even development of new scores might be necessary to reach higher levels of discrimination.

Informations de copyright

Published on behalf of the European Society of Cardiology. All rights reserved. © The Author(s) 2021. For permissions, please email: [e-mail].