Bleeding patients on extracorporeal membrane oxygenation have reduced platelet aggregation and plasma fibrinogen: a longitudinal observational study

This study investigated changes in coagulation and associations with occurrence of bleeding and thrombosis during extracorporeal membrane oxygenation (ECMO) therapy. The study included 100 adult ECMO-patients. Standard coagulation parameters, platelet aggregation and thromboelastometry (ROTEM®) were compared with healthy controls. Data on bleeding and thrombosis were collected until recovery or death. Mortality data were collected 30 days after weaning from ECMO. During ECMO therapy, 53 patients experienced at least one moderate or major bleed. Among these, 42 (79%) patients experienced the first bleeding on day 1 or 2. Platelet aggregation and ROTEM® revealed a hypocoagulable state in ECMO patients when compared with healthy controls. Patients bleeding on day 1 or 2, had lower platelet count (p = 0.04), poorer platelet aggregation and lower levels of fibrinogen (p < 0.01) than patients not bleeding on day 1 or 2. Further, ROTEM® clot propagation was reduced in bleeding patients (p < 0.001). Mortality was higher among bleeding patients than patients not bleeding on day 1 or 2 (67% versus 34%, p < 0.01). Congruity existed between ROTEM® measurements and standard coagulation assays, but plasma fibrinogen had a stronger association with bleeding than ROTEM® measurements. The present study does not support ROTEM® analysis as a routine part of coagulation monitoring during ECMO therapy.

Use of extracorporeal membrane oxygenation (ECMO) has increased within the last decade, particularly the use of ECMO for extracorporeal cardiopulmonary resuscitation (ECPR) in patients with refractory cardiac arrest [1][2][3][4] .Mortality and frequency of bleeding and thrombotic complications vary between ECMO configurations, but bleeding has consistently been associated with mortality 5,6 .Over the past years, the incidence of adverse events in general has decreased, but the incidence of medical bleeding remains unchanged 5 .Non-modifiable risk factors for bleeding or thrombosis, such as age, sex, and weight, have been identified using the Extracorporeal Life Support Organization (ELSO) registry 5,6 , but very limited data exist on the direct impact of coagulation status.
The dynamic process of coagulation from activation of platelets and clotting factors to fibrinolysis can be exhaustively examined only by using a combination of dynamic assays, such as platelet aggregometry and global whole blood coagulation assays (e.g., thromboelastography [TEG] or rotational thromboelastometry [ROTEM ® ]), which are widely implemented in the handling of hemorrhage, especially during liver and cardiac surgery [7][8][9] .
Few studies have evaluated the use of TEG/ROTEM ® during ECMO therapy and small sample sizes hamper clear interpretation of the associations between changes in coagulation and the development of bleeding, thrombotic complications and mortality [10][11][12][13][14][15][16] .
Evaluation of platelet aggregation during ECMO therapy has been performed in few studies [16][17][18] .But as with TEG/ROTEM ® , association between platelet aggregation and bleeding or thrombosis during ECMO therapy remains to be fully elucidated 17 .
ECMO management.All patients were treated at the discretion of the ICU/ECMO team and treatment followed in-house standards closely adapted to ELSO recommendations.ROTEM ® and platelet aggregation results from study blood samples were not available to the attending physician.Patients were cannulated with 17-23 Fr venous cannulas, and 15-21 Fr arterial cannulas, all jugular-femoral (veno-venous (VV)-ECMO) or femoralfemoral (veno-arterial (VA)-ECMO) cannulation.By standard, the ECMO circuit was inspected for fibrin and clot formation by experienced ECMO specialists every 8 h.
Anticoagulation.Patients received unfractionated heparin (UFH) as an intravenous bolus injection of 5000 IU upon ECMO cannula insertion.Heparinization was monitored by measurement of activated partial thromboplastin time (APTT, reference range 20-29 s) at least four times daily; the aim was 40-55 s in VV-ECMO patients and 50-65 s in VA-ECMO patients, obtained by continuous intravenous UFH.Antithrombin levels were kept above 80%.Platelet inhibitors were prescribed at the discretion of the attending cardiologist in collaboration with the ECMO specialist physician.
Transfusions and treatment of bleeding.Decision on platelet transfusions aimed to maintain platelet count above 50 × 10 9 /L.In case of bleeding complications, the transfusion threshold was increased to platelet count of 80-100 × 10 9 /L.The standard threshold for transfusion of red blood cells was a hemoglobin of 6.0 mmol/L (corresponding to 96.7 g/L).
In case of bleeding potentially leading to hypovolemia and subsequent limitations in running sufficient ECMO blood flow, a balanced transfusion strategy was applied with a 1:1:1 ratio of red blood cells, fresh frozen plasma, and platelets.This was supported by 24-h available ROTEM ® guided corrections, adding extra fibrinogen concentrate or fresh frozen plasma if needed.In case of fluid overload, prothrombin complex concentrate was considered and in case of refractory bleeding recombinant factor VII was available.In general, treatment of bleeding event was at the discretion of the attending physicians with special knowledge on ECMO therapy with support from a team of coagulation experts from the Department of Clinical Biochemistry.

Clinical data.
Clinical variables were extracted systematically from the patient medical records and ICU observation charts.These included: ECMO configuration; indication for ECMO therapy; comorbidities selected in accordance with the Charlson Comorbidity Index 19 ; Sequential Organ Failure Assessment (SOFA) Score modified to use the Richmond Agitation-Sedation Scale 20 ; daily evaluation of bleeding or thrombosis events until termination of ECMO support.Bleeding complications were classified according to the Global Utilization of Streptokinase and tPa for Occluded arteries (GUSTO) bleeding criteria 21 and anatomical location.Occurrence of thrombosis was classified as arterial or venous.Further, 30-day mortality after ECMO weaning, transfusion of blood products, and daily therapy with pro-or anticoagulant medications were registered.
Blood sampling and laboratory analyses.Blood for standard coagulation parameters were collected within the daily routine rounds.The first blood sample for platelet aggregation, ROTEM ® analysis or in vivo thrombin generation was collected on the first possible weekday following ECMO initiation, latest on day 3. Blood samples were then collected on weekdays until day 7. Thereafter blood samples were collected on days 14 ± 2 days and 21 ± 2 days, if the patient was still receiving ECMO therapy.Blood was drawn from a non-heparinized arterial cannula already in place.
Healthy controls.For Multiplate ® and ROTEM ® , 80 healthy individuals served as controls (All healthcare workers.For Multiplate ® , two are missing), presented in a previous publication 23 .For TAT measurements, 122 healthy blood donors served as healthy controls and established a reference interval, also previously published 24 .
Markers of organ function and standard coagulation assays.Platelet count, international normalized ratio (INR), APTT, antithrombin, fibrinogen, D-dimer, hemoglobin, and markers of inflammation and organ function were analyzed at the automated routine laboratory according to ISO:15189 accredited protocol.
Data management and statistics.Study data were managed using REDCap electronic data capture tools 25 .Statistical analysis and graphs were performed using GraphPad, Prism 9 (GraphPad Software Inc., CA, USA).All data are presented as median with interquartile range (IQR) for uniformity, as most variables did not follow normal distribution.In figures, data are presented in Tukey plots.To minimize missing data, blood samples were grouped into the following days: day 1 (± 1), day 4 (± 1), day 7 (± 1) and day 14 (± 2).Samples from day 21 (± 2) are not presented as only four patients were still on ECMO for this long.Changes over time are presented graphically, but not tested statistically due to a significant drop-out because of weaning or death.Differences between ECMO patients and healthy controls and between patients with and without bleeding were assessed with unpaired t-tests or with Mann-Whitney test, when appropriate.Distribution of categorical variables between groups was assessed with Fisher's exact test or Chi-square test when appropriate.All tests of significance were two-tailed; p < 0.05 was considered significant.

Results
A total of 100 patients were included for analysis while 24 patients were excluded as they were transferred either from cardiac surgery (post-cardiotomy VA-ECMO) or from other hospitals later than day 3 of their ECMO therapy.
Patient demographics, clinical characteristics, details on ECMO therapy, antiplatelet therapy and mortality are presented in Table 1.In total, 31% of patients received VV-ECMO and 69% VA-ECMO.The main indication for VA-ECMO was ECPR, initiated in 39% of all patients.A total of 778 days of ECMO was evaluated with median duration of 6 (3-10) days.
Level of comorbidity was low according to Charlson's Comorbidity Index and half of the patients were registered as having no comorbidities.Disease severity was stable during ECMO therapy as median SOFA score day 1, 4, 7 and 14 ranged between 12 and 15 (IQRs 9-18).
Markers of organ function and conventional markers of coagulation (INR, APPT, platelet count, antithrombin, fibrinogen, D-dimer) are presented in Table 2. Further, APTT and infusion rate of UFH are depicted.Median and IQR of APTT was within or below target range except for the 75th percentile for VA-ECMO day 1.Over time, infusion rates of UFH doubled while APTT remained stable.

Platelet aggregation.
Measurements of platelet aggregation are presented in Fig. 1.On day 1, platelet aggregation was reduced for all three agonists when compared with healthy controls (all p < 0.0001).The reduction remained significant (all p < 0.0001) when excluding patients, who had a platelet count below 100 × 10 9 /L or who had received either acetylic salicylic acid (ASA) from the ASPItest or ADP inhibitors from the ADPtest.
Over time, median platelet aggregation increased in the ASPItest, possibly due to a decreased use of ASA (Table 1), while the use of ADP inhibitors remained stable as so did platelet aggregation measured by ADPtest.However, platelet aggregation measured by all three agonists remained below the level of the healthy controls (Fig. 1).

ROTEM® and standard coagulation markers.
Results from ROTEM ® analysis are depicted in Fig. 2.
On day 1, ROTEM ® revealed a hypocoagulable profile in ECMO patients in comparison with healthy controls.Clot initiation was prolonged in accordance with the slightly prolonged INR and the prolongation of APTT attributable to the infusion of UFH.The correlation between the 326 pairs of APTT and INTEM CT measurements was significant and moderate, Spearman r s 0.53 (95% CI 0.44-0.60,p < 0.0001).Clot propagation was slower than in healthy controls, while FIBTEM clot firmness was preserved.Median fibrinogen level was within reference range and hence in accordance with the maintained FIBTEM MCF.Platelet MCE was significantly lower in ECMO patients than in healthy controls.In vivo thrombin generation.In vivo thrombin generation was evaluated by measurement of TAT complex levels as shown in Fig. 2. TAT was highest on day 1 with 65% of the patients having levels above our in-house established reference range (≤ 13 µg/L).
Bleeding complications.In total, 66 patients suffered at least one bleeding event during ECMO therapy.
The primary sites of bleeding were from the cannulation sites and from skin or mucosa, found in 52 (79%) patients.Three patients suffered intracerebral hemorrhage.www.nature.com/scientificreports/Bleeding episodes were classified according to GUSTO score; 13 (20%) were minor, 38 (58%) moderate and 15 (23%) severe.The first day of bleeding was day 1 or 2 in 50 (76%) out of the 66 patients, who experienced bleeding.Of these 50 bleedings, 42 (84%) were categorized as moderate or severe.
EXTEM CT and INR did not differ between bleeding and non-bleeding patients (p = 0.74 and p = 0.59), indicating preserved clot initiation.Likewise, TAT levels were comparable between bleeding and non-bleeding patients (p = 0.34).
Patients with moderate or severe bleeding on day 1 or 2 (n = 42) received more blood products on day 1 than patients not bleeding (n = 58) (red blood cells p < 0.001, fresh frozen plasma p < 0.01, platelet concentrate p = 0.001) (Table 3).Few patients received hemostatic medications.During day 1 and 2, one patient received prothrombin complex concentrate (Octaplex ® ), three received fibrinogen concentrate and two patients received vitamin K. None received recombinant activated coagulation factor VII (NovoSeven ® ).

Thrombotic complications.
Fifteen patients were diagnosed with either thrombotic complications (n = 7), clots in the oxygenator (n = 7) or both (n = 1).Seven of the eight patients with thrombotic complications had signs of microthrombosis (critical peripheral ischemia), while one of these patients also had both vena cava inferior thrombosis as well as arterial thrombosis.Five of eight patients had thrombosis detected within the first 2 days, possibly related to infection and sepsis.In four patients, resection or amputation of thrombotic tissue was either planned or performed.
Oxygenator clot was observed 10 times in eight patients during the study period.In four patients, the oxygenator clot resulted in oxygenator change, and in one of these patients, the oxygenator was changed twice.

Mortality.
Of all patients, 48% died, either during ECMO treatment (n = 33) or within 30 days after successful weaning (n = 15); ECPR patients had the highest mortality of 67% (Table 1).In 33% of all patients, ECMO treatment was terminated due to death.Among the patients bleeding on day 1 or 2, more patients died (21/42) than patients not bleeding day 1 or 2 (12/58) (50% vs 21%, p = 0.003, OR 3.8 (95% CI 1.6-8.4)).When counting days of bleeding from day 1 to 7, patients that died had more days of bleeding than patients alive 30 days after

Discussion
The present study reveals a multi-faceted coagulopathy in bleeding ECMO patients.Primarily, platelet aggregation is reduced, but also clot propagation and fibrinogen levels are lower in bleeding patients than in non-bleeding patients.Bleeding episodes occurred early on during the course of ECMO therapy and these early bleeding episodes were associated with mortality.
The present study provides new essential knowledge of importance for the clinical management as it demonstrates low platelet aggregation to be associated with moderate to severe bleeding on day 1 and 2 after initiation of ECMO therapy, irrespectively of platelet count above or below 100 × 10 9 /L.Patients with platelet count < 100 × 10 9 /L showed a higher prevalence of bleeding, pointing towards platelet count as well as platelet aggregation being implicated in bleeding during ECMO therapy, particularly in the early phase of ECMO therapy.This association between platelet count < 100 × 10 9 /L, decreased platelet aggregation and the presence of bleeding stresses the need for more focus on platelets during ECMO therapy.Previous studies revealed low platelet Table 3. Dynamic and standard coagulation parameters in patients during extracorporeal membrane oxygenation therapy divided according to severe or moderate bleeding on day 1 or 2 (n = 42) versus minor or no bleeding day 1 or 2 (n = 58).All data presented as median (interquartile range).Significant values are in bold.*ROTEM ® and Multiplate analysis was performed in 36 bleeding and 51 non-bleeding patients.AUC area under the curve, ADP adenosine diphosphate, ASPI arachidonic acid induced aggregation, TRAP thrombinreceptor-agonist-peptide-6, CT clotting time, MCF maximum clot firmness, MAXVEL maximum velocity, t-MAXVEL time to MAXVEL, MCE maximum clot elasticity, INR international normalized ratio, APTT activated partial thrombin time, TAT thrombin-antithrombin complex.www.nature.com/scientificreports/count 17,26 , acquired von Willebrand disease within the first day of therapy 27 , and decreased platelet aggregation during ECMO therapy 18,[27][28][29] .Tauber et al. found decreased platelet aggregation to be associated with a higher need for transfusion of red blood cells 28 .Transfusion of red blood cells may, however, depend on other factors than bleeding, e.g., hemolysis, specific transfusion limits, and relative ischemia.In another study, Siegel et al. evaluated 30 ECMO patients and found decreased platelet aggregation within 3 days of treatment to be associated with death, but associations with bleeding were not presented and patients receiving platelet inhibitors were included in the analysis.The data in the present study were analyzed with and without patients receiving platelet inhibitors.With regard to global hemostasis ECMO patients, all ROTEM ® parameters changed in a hypocoagulable direction, except for clot firmness in the FIBTEM assay.Bleeding patients had slower clot propagation and lower clot firmness than non-bleeding patients, but clot firmness results remained within the reference range.A larger proportion of bleeding patients than non-bleeding patients had plasma fibrinogen levels below reference range.This is in accordance with a smaller study by Laine et al. 15 , where lower plasma fibrinogen was reported in severely bleeding patients than in non-bleeding patients, although not significantly.Hence, the present study strengthens the indication of low plasma fibrinogen level as an additional risk factor for bleeding in ECMO patients and challenges previous levels of substitution suggested by others, going as low as 150 mg/dL or FIBTEM MCF < 8 mm 16 .Overall, congruity was shown between standard coagulation tests and ROTEM ® measurements.However, plasma fibrinogen was stronger associated with bleeding than FIBTEM clot firmness and, confirming previous studies 13,30 , the correlation between all APTT and INTEM CT measurements was only moderate.Hence, the present study does not support ROTEM® as a routine part of coagulation monitoring during ECMO therapy, but strengthens the indication for measuring plasma fibrinogen along with standard coagulation parameters (INR, APTT, platelet count) when monitoring hemostasis in ECMO patients.
The prevalence of moderate and severe bleeding episodes in the present study was comparable to other studies as summarized by Jiritano et al. 26 .The present study also confirms that bleeding is an early event during ECMO treatment and the risk of bleeding after day 3 is low, as found by others 31 .Of importance, bleeding during ECMO is associated with mortality, which is demonstrated by the present and other studies 5,6 .
The present study only detected symptomatic thrombotic events.Thrombosis or oxygenator clot were registered in 15% of the patients, which is somewhat lower than found in register studies 5,6 .Generally, thrombotic events in ECMO patients are possibly highly underestimated by clinical evaluation 32 .
The strengths of the present study comprise a large cohort compared with previous studies on bleeding and thrombosis during ECMO therapy and with similar patient characteristics in particular concerning age, predominantly male sex, comorbidities, duration, and indications for ECMO therapy [4][5][6] .Patients in the present study were extensively examined with conventional and dynamic coagulation assays evaluating primary and secondary hemostasis including in vivo thrombin generation.Evaluation of platelet aggregation measures were performed with and without exclusion of patients receiving antiplatelet medicine and platelet count above and below 100 × 10 9 /L, as platelet aggregation is negatively affected by platelet count < 100 × 10 9 /L 33 Further, patients were followed during their entire ECMO run and until 30 days after weaning.Based on detailed clinical information the association with bleeding and thrombosis could be investigated.
This study also has limitations.A cause-and-effect relationship could not be established as coagulation measurements and bleeding events were concurrent.Registration of bleeding and thrombosis relied upon medical charts leading to possible underreporting.The registration and grading of bleedings were however, supported by systematic registrations of transfusion requirements and hemodynamic data.The present study did not perform ultrasound following decannulation.Hence, incidental deep venous thrombosis at cannulation sites were not found and these constitute a large part of thrombotic events in previous reports 32 .Due to high early mortality causing drop-out and heterogeneity in the course of disease, repeated measurements analysis was not feasible.
In conclusion, low fibrinogen levels and decreased platelet aggregation, irrespective of platelet count, were associated with bleeding on day 1-2 after ECMO initiation.Congruity existed between ROTEM® measurements and standard coagulation assays, but plasma fibrinogen measurements had a stronger association with bleeding than ROTEM ® measurements and the correlation between APTT and INTEM CT measurements was only moderate.The present study does not support ROTEM ® analysis as a routine part of coagulation monitoring during ECMO therapy.

Table 1 .
Demographic and clinical data on 100 patients receiving extracorporeal membrane oxygenation therapy.Data are provided as median [interquartile range] or n (%) as appropriate.IQR interquartile range, n number, ECMO extracorporeal membrane oxygenation, VV venous-venous, VA venous-arterial, ECPR extracorporeal cardiopulmonary resuscitation, SOFA sequential organ failure assessment score.