Argatroban versus heparin in patients without heparin-induced thrombocytopenia during venovenous extracorporeal membrane oxygenation: a propensity-score matched study

This analysis included all patients with severe respiratory failure (PaO₂/FiO₂ < 85 mmHg or refractory respiratory acidosis with pH < 7.25 on optimized PEEP, usually ≥ 15 cmH₂O) of the Regensburg ECMO-Registry who had received vvECMO in an intensive care unit of the University Medical Center Regensburg between August 2006 and March 2019. In September 2017, the standard anticoagulation regime was switched from UFH to Argatroban (Fig. 1). All patients who had received UFH and had shown an unexplained decline in platelets of ≥ 30% after more than four days of ECMO therapy were routinely screened for HIT. Patients with HIT confirmed by ELISA (antibodies against platelet factor 4; HemosIL Acustar HIT-IgG, Werfen, Germany) and HIPA (heparin-induced platelet aggregation; Department of Immunological and Transfusion Medicine, University Hospital Greifswald) were excluded from the analysis.

Routine data such as demographics, daily laboratory parameters, and disease scores were extracted from the electronic patient data management system of our hospital. The total doses of anticoagulation were recorded. Disseminated intravascular coagulation (DIC) was defined according to the International Society on Thrombosis and Hemostasis DIC score [13]. Further details on the DIC score are provided in the supplement (Additional file 1: Table e1) and have been published previously [4].

The study protocol was reviewed and approved by the local institutional Ethics Committee (19-1335-104). The study was conducted according to the Declaration of Helsinki on Good Clinical Practice.

Before September 2017, UFH was used according to our institutional standard anticoagulation protocol for patients on vvECMO support. After September 2017, Argatroban was used as a first line anticoagulation strategy with an initial infusion rate of 0.2 µg/kg/min [8]. Target activated partial thromboplastin time (aPTT) was 50 ± 5 s in both groups [8, 12]. Argatroban concentration was measured once daily in the Argatroban group. Further details on the anticoagulation strategy are provided in the supplemental material (Additional file 1: Tables e2–4).

The primary end point was the occurrence of major bleeding and/or major thrombosis. Major bleeding was defined according to the recommendations by the ELSO [5] as a drop in hemoglobin of ≥ 2 g/dl/day or in transfusion of ≥ 2 packed red cells/24 h, or retroperitoneal, cerebral, or pulmonary bleeding. Minor bleeding was defined as < 2 packed red cells/24 h. In addition, physicians’ documentation (once per day) and nurses’ shift reports (3 times per day) were evaluated for bleeding events.

Deep vein thrombosis in the area of cannulation (incompressibility of the vein and absence or reduction of flow) was diagnosed by specially trained physicians by means of duplex sonography after decannulation [14, 15]. Obstruction of > 50% of the vessel lumen diameter was classified as major thrombosis and obstruction of ≤ 50% as minor thrombosis [4]. All CT scans, which were performed for clinical reasons during or after ECMO, were assessed for pulmonary embolism.

Secondary end points included major thrombosis, major bleeding, the time-course of the platelet counts, the cumulative transfusion of blood products (red blood cells [RBC], fresh frozen plasma [FFP], and platelets), and the incidence of intracranial bleeding. Cerebral hemorrhage was assessed by means of a CT scan. The absolute number of platelets was documented daily for the entire duration of ECMO support. Additional secondary endpoints were technical and clinical complications such as thrombosis of the membrane oxygenator or pump head, occurrence of hyperfibrinolysis requiring oxygenator exchange according to the criteria previously defined by the ELSO [16], as well as the direct costs of drugs, blood products, and HIT diagnostics. Further information on cost analysis is provided in the supplements (Additional file 1: Table e5).

Propensity-score matching was carried out by matching every patient who received Argatroban with a historical UFH patient to control for potential confounders such as age, sex, body mass index (BMI), sepsis-related organ failure assessment (SOFA), the acute respiratory distress syndrome (ARDS) subgroup (pulmonary, non-pulmonary, post-trauma, and miscellaneous), resuscitation and days of mechanical ventilation before ECMO, and days on ECMO to account for potential imbalances between the Argatroban and the UFH group. More details are presented in the supplements.

All non-normally distributed quantitative data are expressed as median (interquartile range) and were compared with the Mann–Whitney-U test. Differences between the two study groups were assessed with the Chi-squared test of independence for nominal variables. Linear regression models were calculated to assess the association between the Argatroban plasma concentration and aPTT or SOFA. All reported p values were two-sided, and a p value of ≤ 0.05 was considered statistically significant. Data entry and calculation were done using Microsoft EXCEL365 ProPlus (Microsoft, USA), IBM SPSS Statistic software version 25.0 (SPSS Inc. Chicago, IL, USA), and the software package R (v 3.6.1 R Foundation for Statistical Computing).

From September 2017 to March 2019, 43 patients on vvECMO support had received Argatroban for anticoagulation. Four patients had to be excluded from analysis due to receiving a second anticoagulant during ECMO support; thus, our Argatroban group included 39 patients. Our control group included 465 patients on vvECMO support (January 2006 to August 2017) who had been treated with UFH as primary anticoagulant. After propensity-score matching, the control group consisted of 78 patients receiving UFH (Fig. 1). The patient characteristics of the entire study cohort after propensity-score matching are presented in Table 1. Treatment with either Argatroban or UFH resulted in similar rates for renal replacement therapy (Argatroban 41% vs. UFH 40%, p = 0.405), immunosuppression (Argatroban 15% vs. UFH 18%, p = 0.728), and disseminated intravascular coagulation (Argatroban 3% vs. UFH 8%, p = 0.422). Further details and baseline respiratory parameters are provided in the supplement (Additional file 1: Table e7).

In the primary analysis, major bleeding and/or major thrombosis occurred in 31 (79%) patients in the Argatroban group and in 65 (83%) in the UFH group (Additional file 1: Fig. e1). Non-inferiority was shown for Argatroban (upper limit of the 90%-CI < 10%) with an absolute risk difference of -4.9% (90%-CI − 15.6 to 7.9%; p = 0.026).

Anticoagulation during vvECMO therapy is much scrutinized because the device per se may cause both clotting and hemorrhagic complications [16]. Systemic anticoagulation is necessary to lessen potential lethal complications such as thromboembolism [4] but also increases the risk of bleeding [17, 18]. The ELSO anticoagulation guideline does not include any specific recommendations on anticoagulation during vvECMO support [19]. According to a recent survey, 264 out of 273 ECMO centers routinely use UFH as primary anticoagulant [20].

In the current analysis, Argatroban and UFH did not differ regarding the composite endpoint of major bleeding and major thrombosis. Although bleeding complications are common during ECMO support, the reported incidence varies between 0 and 91% [18, 21]. Besides potential differences in treatment algorithms, these variations are most likely due to differing definitions and methods for assessing bleeding [5, 18, 21]. In our study, the rate of bleeding complications was numerically relatively high but was mainly related to the chosen definition of bleeding according to the ELSO classification [5] that is liberal compared to other definitions [22, 23]. Considering only patients with either a drop in hemoglobin of ≥ 2 g/dl/day or a transfusion of ≥ 2 RBC/day, the incidence of major bleeding in our study was below 30%. The average number of RBC transfusions during ECMO support amounted to 0.3/day. Compared with data from a recent meta-analysis including 18 studies, this number is in the lower range [18]. In line with our results, Mazzeffi et al. reported that only the number of transfused blood products but not the bleeding episodes per se were associated with mortality during ECMO support [24]. Also, the incidence of cerebral hemorrhage in both our cohorts was low compared to other reports [25, 26]. The occurrence of bleeding, among others, depends on the pursued level of anticoagulation. Both our patient groups showed similar median aPTT levels of about 50 s throughout ECMO support. To accomplish this level, more aPTT checks per day were necessary in the Argatroban group. In our study, Argatroban doses had to be increased if markers for sepsis were dropping. We did not observe any clear threshold for aPTT or Argatroban-plasma concentration at which the rate of complications increased.

Severe bleeding with suggested standard dosing (2 µg/kg/min) of argatroban has been reported in critically ill patients on vvECMO [8]. Therefore, equally to others, initiating dose was set to 0.2 µg/kg/min in this study [8, 12]. In general, bleeding events are not uncommon during ECMO therapy [24]. Thus, the use of argatroban might be of advantage due to the shorter half-life time compared to UFH but in contrast to UFH an antidot is not available and the choice of how to monitor argatroban dosing is controversial [27]. Still, aPTT is commonly used for controlling anticoagulation with Argatroban [8, 11, 12, 28] and it also was shown that the target range of aPTT in vvECMO patients was more frequently achieved with argatroban compared to UFH [12].

Interestingly, the decline in the platelet count was more pronounced in the UFH group. This often described phenomenon of decreasing platelet numbers [5, 9] seems to be related to the ECMO device, possibly the pump unit, itself [29]. Whether heparin has an additional, independent effect on platelets—which may not occur with Argatroban—needs to be investigated in more detail, in particular due to the numerical inclusion imbalances between the Argatroban and the UFH group.

Besides bleeding, other complications with an important impact on outcome are thromboembolism [4, 30] and technical issues of the ECMO device caused by clotting. The incidence of thrombosis in patients requiring vvECMO support ranges from 0 to 85% [4]. Most previous studies reporting on the use of Argatroban during ECMO therapy included patients with type II HIT [8, 12]. However, HIT is associated with high rates of thromboembolism and hence lower survival rates [10]; therefore, a direct comparison with the current analysis that excludes patients with HIT is not feasible. The largest study on thromboembolism during vvECMO therapy to date reports incidences of any type of thrombosis of 62% [4], and the fraction of 28% clinically significant thromboembolic events was comparable to that of the Argatroban group in this study. Eventually, Argatroban may be an alternative in cohorts with a high risk for thromboembolism such as COVID-19 patients [31] due to its antithrombotic action, presumably also working in already formed clots, and its potential anti-inflammatory, and theoretical antiviral characteristics [32]. Technical complications related to clotting were similar. In 2004, Young et al. [33] found decreased thrombin generation in ECMO patients anticoagulated with Argatroban in comparison to UFH, but this observation has not been transferred into clinical practice so far.

It is likely that aPTT is not the best parameter to monitor anticoagulation with Argatroban and other parameters such as ecarin clotting time may be superior [28]. Yet, aPTT is widely available in clinical practice and has been frequently used to control anticoagulation with Argatroban in ECMO and non-ECMO patients [8, 11, 12, 34]. Alternatives such as Argatroban concentration or plasma-diluted thrombin time that are not influenced by various other factors such as D-Dimers and coagulopathy have no ceiling effect like aPTT and remain valid at higher ranges [10, 27, 35, 36]. In the current analysis, Argatroban concentrations only moderately correlated with aPTT levels. However, a therapeutic target range according to clinical efficacy has not yet been defined. In their case report, Kennedy et al. suggested a range of 0.4–1.2 μg/ml but did not provide any further evidence [37]. Technical complications such as pump or oxygenator failure were comparable between the two groups in our analysis and similar to previous reports [38]. By trend, successful discharge from hospital was higher in the Argatroban group; however, our study was not powered for outcome.

Argatroban has been used for anticoagulation in patients with heparin-induced thrombocytopenia since two prospective studies in patients without ECMO showed a reduction in the composite endpoint (thrombosis, amputation, and death) in comparison to historical controls [11, 39]. Data from a retrospective study on patients with HIT showed a reduction in overall treatment costs despite the fact that absolute drug costs of Argatroban are 17 times higher than those of UFH, mainly because of reduced costs of transfusions [40]. Coughlin et al. [6] reported that overall costs of DTI are comparable to those of UFH after accounting for complications and additional tests. Likewise, the current study showed significantly higher direct drug costs in the Argatroban group, but costs for the two groups were aligning when including transfusions and HIT-testing. Thus, from an economical point of view, our results showed that Argatroban can be viewed as an alternative anticoagulant during vvECMO therapy, and not only for patients with HIT.

This single-center retrospective study was conducted by staff with considerable experience in the use of Argatroban. Therefore, some singularities of Argatroban were known before the initiation of the study, for instance the need of a very low initiation dose and a stepwise increase during the further treatment course. A direct causal relationship cannot be inferred due to the study’s design. Patients with cardiogenic shock during veno-arterial (va) ECMO therapy were excluded a priori to avoid cases of liver failure, for which Argatroban is contraindicated. A systematic bias may have occurred since the classification of bleeding events is limited in retrospective studies. To address this issue, we also reported the daily need of RBC transfusion. Our study included patients from a time period of 13 years. We cannot exclude that some details of management had changed over time; however, standards of care, persons in charge and technical set-up remained essentially the same so that outcome data are comparable to results previously published by our research group.

Finally, due to the limited number of patients treated with Argatroban, small effects and differences may have been missed, and future prospective studies should involve larger cohorts.