Critical risks of haemoadsorption for COVID-19 patients and directions for future evaluations: a nationwide propensity score matched cohort study

Abstract

Haemoadsorption has been suggested as treatment adjunct for sepsis and septic shock, cardiac surgery, acute respiratory distress syndrome, and coronavirus disease 2019 (COVID-19). Randomised clinical trials did not provide conclusive evidence for benefits and even suggest risks in COVID-19 patients. Retrospective observational cohort study based on hospital remuneration data from all COVID-19 patients treated in intensive care units in Germany between 01/01/2020 and 12/31/2021. Regression modelling was performed for 1:1 propensity score matching of 2058 patients. Two-sided probability values for group comparisons and regression models with spline functions controlling for non-linear relationships and medically relevant interaction variables were calculated. In-hospital mortality of patients supported with haemoadsorption was significantly higher compared to matched control patients (74.6% vs. 70.3%, p = 0.0299). Haemoadsorption was associated with coagulopathy (68.0% vs. 54.9%, p < 0.0001), cardiac arrhythmia (49.2% vs. 44.2%, p = 0.0272), and cardiopulmonary resuscitation (CPR, 19.3% vs. 13.1%, p = 0.0002). Further, haemoadsorption increased the chance of death for COVID-19 patients without septic shock (odds ratio, OR [within a 95% confidence interval, CI]; 1.40 [1.05-1.86]) and did not improve survival of septic shock patients (1.19 [0.85-1.67]). Independent variables with a significant impact on mortality included the use of extracorporeal membrane oxygenation (ECMO, 2.15 [1.68-2.76]) and CPR (1.60 [1.03-2.45]). The timing of the haemoadsorption therapy had no effect on patients´ outcomes. Due to inconclusive evidence for benefit and potential harm, haemoadsorption therapy should be limited to thoroughly designed clinical trials before introduced into clinical routine in the context of COVID-19.

Keywords: Acute respiratory distress syndrome; COVID-19; Cytokine adsorption; Haemoadsorption; Hyperinflammation; In-hospital mortality; Intensive care unit; Septic shock.

Fig. 1

Flowchart for patient inclusion in the haemoadsorption and control groups. Each patient receiving haemoadsorption was paired with a control patient who had the most similar propensity score, representing the estimated probability of receiving haemoadsorption based on independent variables.

Fig. 2

Haemoadsorption is a significant risk factor for death in the multiple logistic regression model. Obesity, congestive heart failure, gender, chronic pulmonary disease, renal disease, diabetes, and stroke were statistically insignificant parameters for mortality within the study group but were included in the model due to AIC improvement. Interaction variables show OR for death from (i) septic shock in the control group (septic shock), (ii) Septic shock in the haemoadsorption group (septic shock I haemoadsorption), (iii) Haemoadsorption in the non-septic shock group (haemoadsorption), and (iv) Haemoadsorption in the septic shock group (haemoadsorption I septic shock). ECMO = extracorporeal membrane oxygenation, CPR = cardiopulmonary resuscitation, w/o = without.

Fig. 3

The timing of therapy initiation has no effect on survival in the multiple logistic regression model within the haemoadsorption group. The multiple logistic regression model for the haemoadsorption group included a spline function for the time of therapy initiation after ICU admission to account for the non-linear distribution of patients’ treatment onset, which had no significant impact on mortality. Statistically insignificant parameters that were included in the model with a spline function for timing of therapy initiation after ICU admission due to AIC improvement are three doses versus one dose of haemoadsorption (0.75 [0.51–1.10]), stroke (0.38 [0.14–1.02]), and myocardial infarction (1.85 [0.57–5.99]). ECMO = extracorporeal membrane oxygenation, CPR = cardiopulmonary resuscitation.