Blood exposure to the foreign surface of the ECMO circuit generates an inflammatory response with concomitant activation of the coagulation pathway. Bleeding and thromboembolic complications remain critical issues affecting the outcome of patients undergoing MCS. There is significant variability in the need for anticoagulation according to the device used. The selection of drugs and their dosage is related to the type of device, patient-specific factors, length of treatment and the experience of the medical team . Although heparin remains the most widely used anticoagulant, its pharmacokinetics can be unpredictable with a nonlinear and variable effect. Heparin binds to AT III to inactivate factors IIa and Xa, but the complex heparin-AT III will not inhibit thrombin already bound to fibrin making it ineffective against pre-existing clots [7, 8, 12]. AT III deficiency and the onset of HIT during treatment are serious events, which affect outcome. Although it is argued that the incidence of HIT in ECMO patients is low , its consequences are significant. Patients on V-A ECMO are more likely to experience severe thrombocytopenia and arterial thromboembolism; those on V-V ECMO are more likely to require device or circuit exchange due to oxygenator thromboembolism . Direct thrombin inhibitors (DTIs) have received significant attention in recent years with preference towards argatroban and bivalirudin [11, 12, 33]. Bivalirudin has a half life of approximately 25 minutes, which may be a limitation in areas of blood stagnation, especially during V-A ECMO with non-pulsatile flow . Instead argatroban has a half-life of approximately 45 minutes, which makes it a better candidate as an alternative anticoagulant agent. Furthermore, its pharmacokinetic profile does not appear to be significantly affected by age or gender .
Argatroban and liver function
Argatroban undergoes liver dependent metabolism with four different metabolites, one of which possesses approximately 30% of the parent compound’s activity . Results from in vitro observations support the involvement of the hepatic microsomal cytochrome P-450 enzyme: CYP 3A4 and 3A5 in this pathway . Nevertheless, the inhibition of CYP 3A4 and 3A5 did not result in altered argatroban pharmacokinetics in human studies suggesting the involvement of other significant biochemical processes in its hepatic clearance . Critically ill patients often have some degree of liver function impairment, which may have multiple aetiologies, such as decreased cardiac output, redistribution of splanchnic circulation, poor oxygenation, disseminated intravascular coagulation and congestion due to right heart failure. Liver dysfunction is associated with pharmacokinetic changes resulting in a two to three fold half time prolongation for argatroban , which necessitates significant dose reductions in such patients. Multiple studies indicate altered pharmacokinetic profile of argatroban in critically ill patients [21, 39-42]. Saugel et al. found significantly lower average dose requirements in ICU patients with liver dysfunction compared to those without (0.1 vs. 0.31 μg/kg/min)  .
A number of reports assessed the impact of hepatic dysfunction on argatroban dosing requirements in patients on ECLS [21, 43-45]. Dolch et al. reported a nearly 100-fold dose reduction (from 1.6 to 0.02 μg/kg/min) in a young lung transplant patient on V-V ECMO and acute liver dysfunction. The dose reduction resulted in target range aPTT levels (aPTT 45-60s) without any increased rate of bleeding or thromboembolic events . Felli et al. also used substantially reduced initial infusion rates (starting at 0.05 μg/kg/min) in critically ill ECLS patients . Rouge and colleagues applied a dose reduction, albeit of a lesser degree (from 1 to 0.5 μg/kg/min), necessitated by liver impairment in a patient on V-V ECMO .
On the other hand Dingman et al. found an inverse correlation between argatroban dose and disease severity, as reflected by the modified SOFA score, in a cohort of 20 ECLS patients . Although most patients had impaired liver function classified as Child Pugh class B, further analysis of serum bilirubin, which is the hepatic component of the modified SOFA score, did not show correlation with argatroban dosing requirements. Beiderlinden and colleagues also assessed the relationship between argatroban dosing and liver dysfunction in a cohort of 9 V-V ECMO patients with hepatic impairment . They measured Indocyanine Green clearance, which is a validated marker of hepatic perfusion and function  as well as an independent predictor of mortality in ICU patients  . The authors observed no correlation between argatroban dosing and Indocyanine Green clearance . These findings further underline the challenge in obtaining accurate characterization of liver function in the critically ill, by means of trending a single laboratory parameter. Importantly, in the setting of critical illness, hepatic elimination of argatroban may be substantially diminished, even in the face of only moderately altered conventional liver function parameters.
Argatroban and renal function
Renal impairment and the use of continuous renal replacement therapy (CRRT) are very common in patients on ECLS. Renal dysfunction does not significantly alter argatroban clearance [35, 48]. Neither is its elimination influenced by the use of haemodialysis or CRRT [42, 49]. A recent study reported no differential dosing requirements between ICU patient cohorts on ECLS vs. CRRT . Neither was there any differential dosage requirement revealed between patients receiving various CRRT modalities such as sustained low efficiency dialysis vs. continuous veno-venous hemofiltration) .
Anticoagulation targets and monitoring
Most studies utilized aPTT for therapeutic monitoring of argatroban anticoagulation [18-23, 25, 43, 50-53]. On the other hand, some reports used ACT or a combination of ACT and aPTT for the titration of the argatroban effect [26-28, 54-59].
ACT ranges showed substantial variation across the studies included in the qualitative synthesis. Low limits fall between 150-210 s [26, 27] and high limits between 180-230 s [26, 27] (Table 2). In the reviewed literature, a case report by Johnston et al. applied the highest upper limit for target ACT of 400 s and noted no bleeding complications .
The recommended target aPTT for anticoagulation with DTIs in HIT is 1.5 to 3 times the baseline aPTT value [60, 61]. In the reviewed literature, aPTT target ranges for argatroban in ECMO patients show variations within a relatively wide interval. For studies included in the qualitative synthesis, the lower limit fell between 43-70s [19, 21, 23, 25] and the upper limit between 60 and 100s [18, 20, 21, 25]. In published case reports not included in the qualitative synthesis, the lower limit falls between 45 [43, 55] and 80 s  and the higher limit between 60 [20, 21, 43, 52, 54, 57, 58] and 90 s [51, 55, 56, 59]. In summary, most studies appear to target an aPTT corridor in the vicinity of 50-70 s. The optimal target interval may be influenced by various factors such as recent operations, severe thrombocytopenia, the presence of significant bleeding, and recurrent major thromboembolic complications despite target range aPTT. Indeed, a case report by Sin et al. demonstrates that a number of distinct target intervals may be applied throughout the treatment course of a single patient, depending on the prevailing clinical circumstances. The authors of this paper used four different aPTT target intervals through the ICU management course of their patient .
Menk et al. evaluated a cohort of ARDS patients on V-V ECMO or pumpless Extracorporeal Lung Assist (pECLA) receiving argatroban. The authors found no correlation between bleeding and the maximum aPTT value or the number of aPTT values above 75s. Neither was there any difference with regard to mean aPTT between patients with or without bleeding complications. However, two thirds of bleeding events were associated with maximum aPTT values above 75 s. In the same study, the incidence of thromboembolic events was low, though practically all thromboembolic events occurred when minimal aPTT value were below 50 . These observations further support the legitimacy of choosing an aPTT target corridor falling in the range of 50-70 s. With regard to controllability of target range anticoagulation in ECMO patients, Menk and colleagues noted more frequent dose adjustment requirements during the first two days following argatroban therapy initiation compared to UFH. The number of dose adjustments substantially decreased over time for argatroban but less so for UFH. Furthermore, significantly more sub-therapeutic levels were noted in the UFH group . Cho et al observed shorter time to reach aPTT goal in argatroban treated ECMO patients compared to a control group anticoagulated with UFH (5 vs. 7 hours respectively) . They also found a higher percentage of target-range aPTT values in the argatroban treated cohort compared to the control group . These findings suggest that adequate titration of argatroban anticoagulation is not more challenging than anticoagulation using UFH.
Besides argatroban, a number of additional confounders, typically encountered in a critical care setting, may cause aPTT prolongation: haemodilution, alterations in the level of clotting factors, disseminated intravascular coagulation, antiphospholipid antibodies to name only a few. Thus, aPTT values during argatroban therapeutic monitoring should be interpreted with caution and thorough consideration given to the complete clinical picture.
The Ecarin Chromogenic Assay (ECA) is viewed as a highly specific assay for monitoring Direct Thrombin Inhibitors. It has a linear dose response curve rendering it suitable for usage as a proxy measurement of Direct Thrombin Inhibitor drug levels in blood. Seidel et al. reported no correlation between aPTT and argatroban levels measured by ECA. In this study, approximately two thirds of patients were found to be in the therapeutic aPTT range (45-85 s) while only 9 % showed target argatroban blood levels by ECA (0.5-1.5 μg/ml), with most patients falling below the therapeutic ECA range . No information was available, whether ECMO was used in this particular cohort. The findings are in agreement with the observation by Smythe et al. who reported normal coagulation profile by thromboelastography (TEG) despite aPTT and ACT showing therapeutic range anticoagulation (59 and 240 s, respectively) . These observations may suggest a potential risk for under treatment when using conventional coagulation assays (aPTT, ACT) to guide argatroban therapy. Whether monitoring argatroban effect by ECA would translate to reduced incidence of thromboembolic or bleeding complications remains to be explored.
Several reports suggest that the overall level of critical illness, as reflected by various ICU disease severity scores, is an important determinant of argatroban dosing requirements in ICU patients, both with  and without  ECMO support . This is also consistent with the observations of Begelman et al., who demonstrated a requirement for progressive argatroban dose reduction as the number of failed organ systems increased . Similarly, inverse correlation was shown between argatroban dosing requirements and disease severity scores in patients on ECLS .
The reviewed literature suggests that significant dose reductions are needed compared to the manufacturer recommended initial argatroban dose at 2 μg/kg/min. In a series by Beiderlinden et al., the only patient who received an initial dose of 2 μg/kg/min suffered serious haemorrhagic complications resulting in a dose reduction by a factor of 10 in subsequent patients . The majority of case reports and series utilize a starting dose range between 0.1-0.3 μg/kg/min. Loading dose is usually not utilized except for occasional reports [25, 56]. When comparing patients on argatroban with or without ECMO support, Dingman and colleagues found no significant difference in argatroban dosing requirements . Furthermore, V-A ECMO patients had a numerically lower first therapeutic argatroban dose compared to V-V ECMO patients (0.309 vs. 0.452 μg/kg/min). However, this did not reach the level of statistical significance (p=0.075). The time required reaching anticoagulation target in ECLS patients with argatroban infusion showed significant inter-patient variations ranging from 4 to 20 hours [18, 21, 25]. Variations in patient characteristics, clinical status, and aPTT targets may account for such differences.
Mortality, length of hospital stay, ICU length of stay, and functional outcome
Argatroban is well tolerated over extended periods, with two studies reporting treatment duration exceeding 80 days [43, 45]. To date, argatroban administration has not been directly linked to increased mortality in ECMO patients [19, 25, 28, 29]. In fact, to our knowledge no case report has suggested argatroban treatment as a major culprit for mortality. Length of hospital stay and ICU length of stay appear to be independent of the type of anticoagulation [25, 28, 29]. One report assessed functional outcomes of patients treated with argatroban compared to those managed with UFH and found no difference .
Several studies demonstrated no difference in terms of major bleeding episodes  or transfusion requirements between ECMO patients with or without argatroban anticoagulation [19, 30]. Kawada et al. observed decreased perioperative bleeding in patients undergoing aortic surgery using left heart bypass with argatroban anticoagulation compared to a control group managed on UFH. The authors provided some evidence that suppressed thrombin-dependent thrombocyte activation in the argatroban group could contribute to such differential effect . On the other hand, Lubnow et al. reported higher bleeding rate in patients treated at least temporarily with argatroban compared to the ECMO control group managed with UFH . In this context, it is important to point out that argatroban treatment was started on clinical suspicion of HIT. One of the hallmarks of HIT is thrombocytopenia, which in itself may result in increased bleeding risk. Indeed, Neissen and colleagues reported a higher rate of bleeding in patients with thrombocytopenia and lower rates were found after argatroban treatment implementation . Dingman et al. noted higher rates of bleeding events and transfusions requirements in argatroban treated patients on ECLS compared to argatroban treated patients without ECLS, which is an expected finding, given the increased rate of haemorrhagic complications associated with the use of ECLS  . Some reports note major bleeding episodes in conjunction with argatroban treatment, which occurred in a perioperative context [23, 64] or at dosages substantially higher than the usually applied range in ICU practice . Available data in the literature on reversal of argatroban effect is scarce. Successful reversal of residual argatroban effect-related bleeding using recombinant factor VII concentrate has been described . Taken together, when applied in adequate doses argatroban is not associated with an elevated haemorrhagic risk compared to UFH. Before initiating argatroban anticoagulation, thorough consideration should be given to concomitant factors associated with increased risk of bleeding such as thrombocytopenia, septic coagulopathy and the use of platelet inhibitors.
Patient and circuit thrombosis
Patient thromboembolic complications are an important source of ECMO related morbidity and mortality. Several studies found no difference in the rate of patient related thromboembolic complications in cohorts treated with argatroban versus UFH [19, 25, 28, 29]. Other studies where no statistical comparison could be made to a control group found an overall low incidence of thromboembolic complications [18, 23, 27].
In general, ECMO system clotting with argatroban anticoagulation was uncommon [18, 21, 26]. Menk et al. reported no difference with regard to circuit clotting between patients on argatroban vs. UFH anticoagulation, except for a subgroup on pump-less ECLA, where the number of clotting events was higher in the argatroban group. This observation may possibly be explained by low flow states prevalent in pump-less systems .
Cost effectiveness plays a major role when opting for a certain therapeutic modality. A recent retrospective study by Cho and colleagues compared the average cost of ECMO course in a cohort of patients anticoagulated with argatroban vs. UFH . The authors found that despite higher drug cost the ECMO course was more cost-effective in the argatroban group compared to the UFH group (7092 vs. 15323 $). Factors included in the cost analysis were drug cost, blood product costs, and costs associated with laboratory tests. The most significant factor accounting for higher cost in the UFH group was the frequent need for AT-III substitution.