Extracorporeal membrane oxygenation survival: External validation of current predictive scoring systems focusing on influenza A etiology
Abstract
Despite increasing clinical experience with extracorporeal membrane oxygenation (ECMO), its optimal indications remain unclear. Here, we externally evaluated all currently available ECMO survival-predicting scoring systems and the APACHE II score in subjects undergoing veno-venous ECMO (VV ECMO) support due to acute respiratory distress syndrome (ARDS) with influenza (IVA) and non-influenza (n-IVA) etiologies.
Our aim was to find the best scoring system for influenza A ARDS ECMO success prediction. Retrospective data were analyzed to assess the abilities of the PRESERVE, RESP, PRESET, ECMOnet, Roch, and APACHE II scores to predict patient outcome.
Patients treated with veno-venous ECMO support for ARDS were divided into two groups: IVA and n-IVA etiologies. Parameters collected within 24 hours before ECMO initiation were used to calculate PRESERVE, RESP, PRESET, ECMOnet, Roch, and APACHE II scores.
Compared to the IVA group, the n-IVA group exhibited significantly higher ICU, 28-day, and 6-month mortality (P = .043, .034, and .047, respectively). Regarding ECMO support success predictions, the area under the receiver operating characteristic curve (AUC) was 0.62 for PRESERVE, 0.44 for RESP, 0.57 for PRESET, and 0.67 for ECMOnet, and 0.62 for Roch calculated for all subjects according to the original papers.
In the IVA group, APACHE II had the best predictive value for ICU, hospital, 28-day, and 6-month mortality (AUC values of 0.73, 0.73, 0.70, and 0.73, respectively). In the n-IVA group, APACHE II was the best predictor of survival in the ICU and hospital (AUC 0.54 and 0.57, respectively).
From all possible ECMO survival scoring systems, the APACHE II score had the best predictive value for VV ECMO subjects with ARDS caused by influenza A-related pneumonia with a cut-off value of about 32 points.