Adsorption therapy in critically ill with septic shock and acute kidney injury: a retrospective and prospective cohort study

After obtaining approval for participation in the Cytosorb^® registry by the local ethics committee [AS 88(bB)/28.01.2015, Chairman Prof. M. Matthias], we prospectively collected data in our ICU at a tertiary hospital in Cottbus, Germany, starting from 2015, for a period of 3 years [20]. We decided to utilise haemoadsorption with Cytosorb™ in all patients in the aforementioned period with following condition: septic shock (rising noradrenaline dose above 20 µg/min.) with sepsis-associated acute kidney injury in adult patients treated in our ICU. The Cytosorb^® registry was designed to explore the use of the Cytosorb™ (Cytosorbents Corp., NJ, USA) absorber in critically ill patients under real-life conditions using all relevant information (e. g., diagnosis, comorbidities, treatment/concomitant medication, clinical laboratory parameters, outcome) during haemoadsorption (ClinicalTrials.gov Identifier: NCT02312024) [20, 21]. The primary endpoint of this prospective, multicentre registry involving more than 130 centres from 22 countries was to compare the in-hospital mortality predicted according to the APACHE II score. APACHE II was calculated for the 24 h prior to initiation of CVVHDF/haemoadsorption (according to the registry protocol). Additionally, we retrospectively analysed identified patients’ group from our ICU treated from 01/2012 to 12/2013 with identical clinical inclusion criteria as a historical control group with approval from the local ethics committee. In detail, inclusion criteria were septic shock (rising noradrenaline dose above 20 µg/min.) with sepsis-associated acute kidney injury in adult patients treated in our ICU. Due to reconstruction measures and the used patient data management systems, we could only extend the historic control group back to 2012. The 3 years of prospective data recruitment were by default from the study protocol of the Cytosorb^® registry. Thus our study was a single-centre prospective intervention study with historical controls.

All patients were treated in accordance with the local sepsis treatment protocol, which was based on the actual guidelines valid during each treatment period [22, 23]. We evaluated the adequacy of fluid resuscitation therapy or vasopressor support using pulse indicated continuous cardiac output (PiCCO^®, Getinge Deutschland GmbH, Rastatt, Germany) or ultrasound. The initial calculated antimicrobial therapy based on the assumed or underlying septic focus according the actual hospital recommendations. The continuous veno-venous haemodiafiltration (CVVHDF, Prismaflex, M150 Filter, Fa. Baxter International Inc.) with or without haemoadsorption was used in all patients with septic shock with acute renal failure (failure stage of RIFLE criteria for 12–24 h, comparative to KDIGO stage 3) according to the RIFLE classification [24], increasing catecholamine dependency above a noradrenaline dose of 20 µg/min (despite adequate fluid resuscitation therapy) and elevated IL-6 higher than 500 pg/ml. All patients were anticoagulated via the continuous administration of heparin. Thus, all patients included in our analysis presented with septic shock, an increasing catecholamine dependency above a noradrenaline dose of 20 µg/min and were treated with CVVHDF due to ARF.

Absorbers were changed when the CVVHDF system had to be changed or further adsorption therapy was needed. The further need for adsorption therapy was assessed whether the patient became increasingly unstable again after termination of the adsorption therapy or interleukins started increasing again.

Starting in January 2015, we prospectively acquired data for the Cytosorb^® registry. According to the underlying protocol of the Cytosorb^® registry, we chose ICU mortality as primary endpoint for our evaluation as well. Secondarily, we evaluated the hospital mortality and length of treatments in days (ICU/hospital/ventilatory support/catecholamine therapy/CVVHDF).

Data analysis and statistics were performed with IBM SPSS Statistics 25. Data are described with median or mean, as necessary. The interquartile range (IQR) or 95% confidence interval is displayed in parentheses as appropriate. Because of the small sample size, all tests were performed as exact tests. To assess group differences, we used Levene’s test and the Mann–Whitney U test. We used α = 5%. Missing data regarded only laboratory test results, and therefore did not influence the interpretation of primary or secondary results. Significant results and differences are marked with an asterisk (*) in the tables. For the purposes of this article, the term “significant” is used when data reached statistical significance, defined as p < 0.05. Because of the missing information in the literature, no power analysis could be performed initially. Therefore, we decided to participate in the CytoSorb^® registry according to protocol. The control group size was also not calculated, but we screened and evaluated all patients since the implementation of our patient data management system in 2012 until the beginning of the Cytosorb™ implementation in 2014. To enable comparison between the groups, we calculated the observed-to-expected (ICU and hospital) mortality ratio (O/E ratio). Further analyses included logistic regression of ICU mortality and hospital mortality adjusted with the covariates "haemoadsorption", "APACHE II", "PCT prior to CVVHDF", "CRP prior to CVVHDF " and "catecholamine dosage before initiation of CVVHDF". The same covariates were used in a Fine and Gray model for the subdistribution hazard ratio of the event of "alive discharge from ICU" in presence of the competing event of death in the ICU.