Patient‐tailored platelet transfusion practices for children supported by extracorporeal membrane oxygenation

Extracorporeal membrane oxygenation (ECMO) serves as cardiopulmonary therapy in critically ill patients with respiratory/heart failure and often necessitates multiple blood product transfusions. The administration of platelet transfusions during ECMO is triggered by the presence or risk of significant bleeding. Most paediatric ECMO programmes follow guidelines that recommend a platelet transfusion threshold of 80–100 × 109/L. To reduce exposure to platelets, we developed a practice to dynamically lower the threshold to ~20 × 109/L. We describe our experience with patient‐tailored platelet thresholds and related bleeding outcomes.


INTRODUCTION
Extracorporeal membrane oxygenation (ECMO) is a life-saving heart-lung bypass modality for patients with severe, reversible cardiac and/or respiratory failure.Although the exact mechanisms are still not fully clear, ECMO induces a reduction in the number of platelets and alteration in platelet function along with activation of complement and coagulation cascades [1][2][3].Thrombocytopaenia is very common in ECMO-supported patients; observational studies in children have found that the platelet count drops by 26%-47% at the time of cannulation [4,5].Abnormal activation of platelets has also been observed in ECMO-supported patients [4].
In order to minimize the risk of bleeding, children on ECMO are transfused large volumes of platelets during their course of ECMO by multiple repeated transfusions.Recent studies have shown that platelets are transfused to children on two-thirds of each ECMO day, and each child on ECMO receives approximately 90 mL/kg of platelets throughout their ECMO course [6][7][8][9][10][11].
Paediatric ECMO-supported patients are commonly anticoagulated and often have surgical incisions, which further increases the risk of bleeding [12][13][14].The challenge of maintaining a balanced haemostatic system has prompted guidelines for the transfusion of blood products on ECMO.The Association for the Advancement of Blood & Biotherapies (AABB), the Extracorporeal Life Support Organization (ELSO) and a recent consensus statement recommend transfusing platelets in a non-bleeding child on ECMO when the count is <80-100 Â 10 9 /L [15][16][17].These guidelines are mainly based on expert opinion but there is a paucity of evidence-based data on the optimal threshold for platelet transfusions in paediatric ECMO patients.Despite the lack of prospective studies, many centres strictly follow these guidelines and set platelet transfusion goals at 80-100 Â 10 9 /L or higher [1,6,9,11,[18][19][20].
We sought to evaluate clinical outcomes, including bleeding complications, in ECMO-supported paediatric patients who were transfused based on a protocol allowing for a gradual decrease in the transfusion threshold to 20 Â 10 9 /L.We hypothesized that the protocol would decrease exposure to platelet transfusions without any increase in the rate of bleeding.

Study design
We conducted a retrospective cohort study of all children and young adults supported by ECMO in the cardiac intensive care unit (CICU) of a tertiary university-affiliated paediatric centre.The study was approved by the Institutional Review Board (Rabin Medical Center approval number 0676-18) who waived informed consent.

Study participants
We included all consecutive paediatric patients and young adults (birth to 28 years of age) with respiratory and/or cardiac failure who were supported by ECMO in our CICU from June 2010 to June 2020.There were no exclusion criteria.

Platelet transfusion practice
With the goal of reducing platelet product use and patient exposure, our practice for the past 10 years has been to individually tailor the platelet transfusion threshold for each patient and gradually lower the threshold from 100 Â 10 9 /L in the first ECMO day to $20 Â 10 9 /L in non-bleeding patients, within 4-5 days of ECMO cannulation.The platelet transfusion threshold was determined during morning rounds for each patient according to age, illness severity and bleeding risk factors (as defined by coagulopathy, hepatic failure, recent surgery or the presence of active or recent bleeding).When bleeding occurred, the patient was transfused with blood products and/or haemostatic medications as needed.When the transfusion of platelets was required, per the pre-defined transfusion threshold, we used apheresis (single donor) platelets at a dose of 20 mL/kg.

Anticoagulation and transfusion management
The patients were anticoagulated with an infusion of unfractionated heparin to maintain an activated clotting time (ACT) of 180-220 s (Hemochron ® Response coagulation monitor, Werfen, Bedford, MA, USA, normal ACT 70-120 s).The heparin infusion was adjusted to a lower ACT (160-180 s) in bleeding patients and to 200-220 s when the ECMO flow was reduced.The complete blood count was assayed daily.
When the patients suffered a major bleeding event, fresh frozen plasma (FFP) was transfused if the ACT or activated partial thromboplastin time was higher than desired (usually >200 or >80 s, respectively), or when no response to tranexamic acid infusion was observed.
If the patient was thrombocytopaenic (i.e., platelet count <100 Â 10 9 / L) at the time of the bleeding, single-donor platelets were transfused.
Antithrombin replacement was accomplished through the transfusion of plasma because antithrombin concentrate is not available in Israel.
Cryoprecipitate was transfused prophylactically when the serum fibrinogen dropped below 100 mg/dL.

Definitions of endpoints and transfusion classification
Major bleeding events were classified as bleeding episodes that met at least one of the following criteria: 1. Intracranial haemorrhage (ICH; excluding intraventricular haemorrhage grade 1) as determined by ultrasound in infants, and CT or MRI imaging in older children.
2. Bleeding episodes that required any of the following: discontinuation of heparin infusion, transfusion of FFP, cryoprecipitate, packed red blood cells (PRBCs) or treatment with tranexamic acid or recombinant factor VIIa.
3. Re-operation for haemostatic control of surgical or cannulation site.
Bleeding events that did not meet these criteria were classified as minor.
Bleeding sites were recorded as the following: central nervous system, pulmonary, gastrointestinal, genitourinary, oropharyngeal, surgical site, cannulation site or post-procedural bleeding.A 'bleeding day' was defined as a day when a bleeding event occurred during the ECMO course, including days when the bleeding continued from a previous bleeding day (the same bleeding event).All bleeding events were reviewed and adjudicated by a senior paediatric cardiac intensivist.
Platelet transfusions were classified as prophylactic, that is, thrombocytopaenia in a non-bleeding patient; therapeutic, that is, active bleeding in a thrombocytopenic patient; minor bleeding in a volumedepleted patient; or pre-procedural.Active bleeding was defined as visually observed ongoing bleeding from mucus membranes, surgical or cannulation sites, gastrointestinal tract, airways and/or acute drop in haemoglobin levels with imaging evidence of ICH.
Survival was recorded at hospital discharge.Follow-up data were extracted only for patients who suffered from ICH while on ECMO.
ICH was screened by head sonography three times weekly.If bleeding was noted, sonography was repeated daily.Grades II-IV intraventricular haemorrhage, as previously classified, were included in ICH bleeding events [21].Screening for ICH in older infants and children was based on daily neurological examinations during interruptions of sedation.This was achieved by temporary discontinuation of muscle relaxants and lightening of sedation regimen until movement of the head, limbs and evidence of respiratory effort.A neurological deficit in a physical examination prompted brain imaging.

Data collection and statistical analysis
Demographic, clinical and ECMO data were abstracted from the electronic health records.Data regarding transfusion of PRBCs, platelets, FFPs, cryoprecipitate and haemostatic medications were recorded, as well as daily platelet counts and bleeding events.
Statistical analysis was carried out using the SAS Software, version 9.4 (SAS Institute, Cary, NC).Continuous variables are presented as medians and interquartile ranges (IQR), and categorical variables by frequencies (percentages).The Mann-Whitney U test was used to compare the lowest platelet counts in the children who bled versus those who did not bleed.In addition, the Spearman rank correlation test was used to determine the relationship between the lowest platelet count and days of bleeding.

Patient characteristics
The analysis included 229 children who met the inclusion criteria; 55% were males and 53% were neonates (age ≤28 days).The majority (94%) were supported by venoarterial ECMO.Demographic, ECMO and outcome data are presented in Table 1.No patient required recannulation after disconnection from the ECMO circuit.

Platelet counts
During the study period, 3735 complete blood counts were assayed, At some point during their ECMO run, 98.7% of the patients were in the thrombocytopaenic range and over half (56.3%) of the patients had a platelet count of <30 Â 10 9 /L at least once during their ECMO run.

Haemorrhagic complications and platelet counts
While supported by ECMO, 140 (61%) patients (95% CI: 55%-67%) experienced bleeding (Figure 4).Twenty (8.7%) patients, including 16 neonates, had ICH.Two patients underwent craniotomy for haematoma evacuation after weaning from ECMO support.On long-term follow-up, one infant developed paraplegia and a seizure  The lowest platelet count of each patient had a weak, but significant negative correlation with the number of bleeding days, that is, the lower the platelet count, the higher the number of bleeding days.
The Spearman rank correlation coefficient between the number of bleeding days and the lowest platelet count was À0.358 ( p < 0.001), as seen in Figure S1.

DISCUSSION
In this retrospective analysis of 229 ECMO-supported paediatric patients, we present our dynamic, patient-tailored, restrictive platelet transfusion practice.As far as we know, this is the first large-scale study that challenges the traditional paradigm of maintaining the platelet count >80-100 Â 10 9 /L in paediatric ECMO-supported patients.We found that while differences were seen in the lowest platelet counts of children who bled on ECMO versus those who did not, the overall median platelet counts were maintained much lower than current recommendations suggest.
Recently, a group of paediatric critical care and haematology experts (TAXI-CAB group) formulated recommendations for the transfusion of platelets and/or plasma in neonates and children undergoing cardiac surgery and/or supported by ECMO [17].The group concluded that evidence to recommend a specific threshold for platelet transfusion is lacking.Specifically, prophylactic transfusion in the absence of clinically significant bleeding seemed unlikely to benefit ECMO-supported patients when the platelet count is >100 Â 10 9 /L.
The guidelines state that most centres maintain platelet levels >100 Â 10 9 /L and emphasize that platelet transfusions may expose patients to risks without potential benefit and should be avoided when possible.These recommendations were incorporated in guidelines that address transfusion policies in critically ill paediatric patients [25].Similarly, in a collaborative study that evaluated the epidemiology of haemostatic transfusions in 514 paediatric patients, the daily median platelet transfusion goal was 100 Â 10 9 /L, with some variability in subgroups of non-bleeding patients, veno-arterial ECMO and non-neonates [11].In a recent survey among 108 medical centres, platelet transfusion thresholds in paediatric ECMO-supported patients were shown to vary by age and centre; 91-100 Â 10 9 /L was the most common threshold for platelet transfusion, and most of the centres without age-based thresholds maintained a threshold >61 Â 10 9 /L [20].
These findings were similar to a survey from 2013, which documented a median platelet transfusion threshold of 100 Â 10 9 /L among 187 ELSO centres [26].There are, however, centres that have adopted a lower threshold than commonly recommended [6,18,19].
In a substantial proportion of our patients, the lowest platelet count was in the severe thrombocytopaenia range (<30 Â 10 9 /L), and almost all the patients were moderately thrombocytopaenic at some point.These counts are significantly lower than found in published data, in which the median platelet counts were close to 100 Â 10 9 /L during the ECMO course [9,11].Although we did demonstrate a difference in the lowest platelet counts in children who bled versus those who did not, all the median counts were much lower than current recommendations.These results provide further evidence that it is safe to set a lower threshold for platelet transfusions than current recommendations suggest; however, there is likely a lower limit below which is associated with a higher bleeding risk.This must be studied in larger, prospective cohorts.
The contribution of thrombocytopaenia to the risk of bleeding has long been debated.The presence and degree of thrombocytopaenia has been shown to have a poor association with bleeding, mortality or ICH in ECMO-supported patients or in those with ICH [11,[27][28][29].
Nellis et al. showed that a platelet count as low as ≤55 Â 10 9 /L for children on ECMO was not associated with increased chest tube output [30].Furthermore, a large randomized trial in premature neonates found increased bleeding and/or mortality and worse 2-year neurological outcomes in those transfused at a higher threshold as compared to a lower threshold [31,32].
In our series, platelets were transfused on only 28.5% of the ECMO days; the median daily platelet transfusion dose was 8.7 mL/kg, and the total median transfusion dose was 70 mL/kg for the entire ECMO run.A recent study investigating the epidemiology of platelet transfusion in the paediatric ECMO population showed that daily platelet transfusions were positively associated with chest tube output, bleeding requiring RBC transfusion, plasma transfusion and set platelet goal but not with platelet counts and that the haemostatic goals did not change according to the bleeding status of the patients.Compared to our study, values were higher for the proportion of platelet transfusion days (67.8%), and the daily median and total median transfusion doses (17.3 and 82 mL/kg/day, respectively) [11].Significantly higher platelet transfusion volume was also observed in a study that investigated 2 anticoagulation protocols (25.4 and 20.1 mL/kg/day) [19].
Practicing a more restrictive platelet transfusion approach did not result in greater PRBC exposure in our patients compared to patients in other studies.This is evident from our median daily PRBC dose of 14 mL/kg/day compared to 21.2-46.1 and 38.7 mL/kg/day previously reported [19,42].It should also be noted that PRBC transfusions in paediatric ECMO are frequently used in non-bleeding days and, on many occasions, when the lowest haematocrit was above the institutional transfusion threshold [42].
The proportion of our patients with any bleeding was 61.1%, which is lower than any previously published series (63.9%-76%), but higher than reported in the ELSO registry report (39%) [9,14,18,43].The major difficulty in comparing bleeding rates in paediatric ECMOsupported patients is the lack of a uniform definition of bleeding.Some studies identified bleeding events based on the need for transfusion, with the exception of ICH, without standardization of transfusion triggers, while ELSO has specific criteria to define major bleeding [14,16].
In addition, while others, including ELSO, record bleeding from the same source that occurs over multiple days as one bleeding day, we considered each day of an ongoing haemorrhage as a separate bleeding day [14].Likewise, oropharyngeal bleeding, which was common (25.8%) in our cohort, could have been regarded as minor bleeding and discarded in other series.None of our patients had significant genitourinary bleeding, in contrast to 6% of the patients in one series [14].
Intracranial haemorrhage was evident in 8.7% of our patients.
This rate is within the reported range by other studies and by the ELSO registry (3.0%-17%) [6,9,13,14,16,31,37,[43][44][45][46][47][48][49][50][51].Comparing ICH rates between studies is challenging because of the lack of standardization of the imaging modality, frequency of testing and severity.As ICH does not usually cause a decrease in haemoglobin level or necessitate blood transfusion, it may have been underdiagnosed in some series.Lastly, an analysis of the ELSO registry documented haemorrhagic deaths (defined as death after withdrawal of ECMO support due to intractable bleeding) in 3.4% of patients compared to none in our cohort [43].
The main limitation of our study is the retrospective design, which could have been subject to recall and misclassification biases.To mitigate these concerns, we examined all the daily nursing and medical reports of the patients for descriptions of bleeding events.A singlecentre report may limit generalizability of our findings but ensures consistency of practices and management protocols, which may be lacking in multi-site studies.The dynamic-transfusion goal was not standardized or recorded in the EMR and hence is lacking from the data.Nonetheless, our platelet counts do demonstrate a lower transfusion threshold than other centres have demonstrated.Another limitation is the lack of timing of platelet transfusions in relation to bleeding or to platelet count measurements.We do not have access to data on donor exposure of the patients within our cohort.As with retrospective studies, we were not able to adjust for confounders when comparing our data to the pertinent literature, which may weakened our comparisons.Many other factors may impact bleeding outcomes, such as the parameters used for anticoagulation, and were not accounted for in the comparison of bleeding events.In addition, data regarding outcomes should be interpreted with caution, as some of our patients were transferred to the referring centre after decannulation, where long-term follow-up data was inaccessible.
In conclusion, this single-centre cohort study showed that a 'patient-tailored' rather than 'goal-directed' platelet transfusion policy is safe and feasible with no apparent excessive blood transfusion or increase in bleeding complications rates.Although our protocol challenges the existing convention, daily adjustment of the transfusion threshold based on a patient's general condition, bleeding tendency and risk factors may prove to better serve our patients.A prospective randomized controlled study examining outcomes of paediatric ECMO-supported patients allocated to different platelet transfusion thresholds is urgently needed.

averaging 1 .
5 counts/patient/ECMO day or 16.3 counts per patient.The overall median (IQR) platelet count was 72 (44-111) Â 10 9 /L.The daily median, IQR and minimal platelet counts of the cohort during the first 28 days of ECMO support are shown in Figure 1.The distribution of the individual minimal platelet counts of the patients is presented in Figure 2.
disorder, and the other was neurologically intact.Three patients died directly from cerebral bleeding.Procedural bleeding (i.e., following tracheostomy or chest drain insertion) occurred in seven (3.1%) patients.The median (IQR) lowest platelet count in children who had bleeding events was 25 (15-32) Â 10 9 /L.This was significantly different from the median (IQR) lowest platelet count in children who did not have bleeding events, which was 33 (23-48) Â 10 9 /L ( p < 0.001), as shown in Figure 5.

1
Median, interquartile range (IQR) and minimal platelet counts during the extracorporeal membrane oxygenation (ECMO) course.F I G U R E 2 Distribution of the patients' minimal platelet counts.F I G U R E 3 Proportion of patients who received platelet transfusions over the course of the (ECMO) run.F I G U R E 4 Rates of bleeding days and haemorrhagic complications in our cohort (filled circles) and in the pertinent literature (empty circles).Reference numbers appear in parentheses.ECMO, extracorporeal membrane oxygenation; ICH, intracranial haemorrhage.F I G U R E 5 Boxplot of lowest platelets counts by bleeding status.
T A B L E 1 Demographic and ECMO data.