TRANSFUSION PRACTICE

Red blood cell transfusion in critically ill children

Pierre Demaret, Marisa Tucci, Thierry Ducruet, Helen Trottier, and Jacques Lacroix

A
pproximately 15 million red blood cell (RBC)
BACKGROUND: Red blood cell (RBC) transfusions are units are transfused annually in the United
common in the pediatric intensive care unit (PICU). States and approximately 85 million world-
However, there are no recent data on transfusion prac- wide.1 RBCs are transfused to increase hemo-
tices in the PICU. Our objective was to determine trans- globin (Hb) concentration and oxygen delivery (DO2),
fusion practice in the PICU, to compare this practice which should increase cellular oxygen consumption.
with that observed 10 years earlier, and to estimate the However, benefits of RBC transfusion must be weighed
compliance to the recommendation of a large random- against risks. Studies have described an association in
ized clinical trial, the Transfusion Requirements in Pedi- intensive care unit patients between RBC transfusions and
atric Intensive Care Unit (TRIPICU) study. adverse outcomes such as mortality,2-5 length of stay and
STUDY DESIGN AND METHODS: This was a single- of mechanical ventilation,2,4-6 infection,4,7 and transfusion
center prospective observational study over a 1-year reactions.8-10 Because the risks of transfusions may out-
period. Information was abstracted from medical charts. weigh their benefits, it is important to analyze transfusion
Determinants of transfusion were searched for daily practice to improve it if weaknesses are identified.
until the first transfusion in transfused cases or until RBC transfusions are frequent in the pediatric inten-
PICU discharge in nontransfused cases. The justifica- sive care unit (PICU). Twelve years ago, 14.1% of children
tions for transfusions were assessed using a admitted to our PICU received at least one RBC transfu-
questionnaire. sion.11 More recently, Bateman and coworkers4 found that
RESULTS: Of 913 consecutive admissions, 842 were 49% of children staying in the PICU more than 2 days were
included. At least one RBC transfusion was given in transfused at least once. Stated transfusion practice
144 patients (17.1%). The mean hemoglobin (Hb) level
before the first transfusion was 77.3 ⫾ 27.2 g/L. The
determinants of a first transfusion event retained in the ABBREVIATIONS: AUC = area under the curve;
multivariate analysis were young age (<12 months), MODS = multiple organ dysfunction syndrome;
congenital cardiopathy, lowest Hb level of not more nTC(s) = nontransfused case(s); PELOD = Pediatric Logistic
than 70 g/L, severity of illness, and some organ dys- Organ Dysfunction; PICU = pediatric intensive care unit;
functions. The three most frequently quoted justifica- PRISM = Pediatric Risk of Mortality; ROC = receiver operating
tions for RBC transfusion were a low Hb level, intent to characteristic; TC(s) = transfused case(s); TRICC = Transfusion
improve oxygen delivery, and hemodynamic instability. Requirements in Critical Care; TRIPICU = Transfusion
The main recommendation of the TRIPICU study was Requirements in Pediatric Intensive Care Unit.
applied in 96.4% of the first transfusion events.
CONCLUSIONS: RBC transfusions are frequent in the From the Division of Pediatric Critical Care Medicine,
PICU. Young age, congenital heart disease, low Hb Department of Pediatrics, the Research Center, and the
level, severity of illness, and some organ dysfunctions Department of Social & Preventive Medicine, Research Center,
are significant determinants of RBC transfusions in the Sainte-Justine Hospital and Université de Montréal, Montreal,
PICU. Most first transfusion events were prescribed Quebec, Canada.
according to recent recommendations. Address reprint requests to: Jacques Lacroix, MD, CHU
Sainte-Justine, Room 3431, 3175, Chemin de la
Côte-Sainte-Catherine, Montréal QC, H3T 1C5, Canada; e-mail:
[email protected].
Supported by Fonds de la Recherche en Santé du Québec
(Grant 24460).
Received for publication May 30, 2012; revision received
April 12, 2013, and accepted April 12, 2013.
doi: 10.1111/trf.12261
TRANSFUSION 2014;54:365-375.

Volume 54, February 2014 TRANSFUSION 365

DEMARET ET AL.

patterns show significant variation among pediatric Cases were defined as transfused cases (TCs) if at
intensivists, as documented by two surveys,12,13 showing least one RBC transfusion was given during the PICU stay
that the threshold Hb for RBC transfusion varied from 70 and as nontransfused cases (nTCs) if no RBC transfusion
to 120 g/L in stable patients with similar diseases. was given.
These studies were conducted while there was no
evidence-based support regarding transfusion strategies
in the PICU. The Transfusion Requirements in PICU Data collection and management
(TRIPICU) study published in 200714 showed that a Hb Trained research assistants collected data daily in a vali-
threshold of 70 g/L for RBC transfusion can decrease dated case report form. All information was abstracted
transfusion requirements in stable critically ill children prospectively from medical charts. If a patient was read-
without increasing adverse outcomes compared to a Hb mitted within 24 hours of PICU discharge, both PICU stays
threshold of 95 g/L. were considered as the same stay. Any readmission occur-
While a randomized controlled trial (RCT) like ring more than 24 hours after a prior PICU discharge was
TRIPICU generates compelling evidence, this does not considered a separate admission.
imply that knowledge application occurs immediately.15 It Baseline data collected within 24 hours after PICU
may take several years before new knowledge is applied at entry included age, sex, weight, and admission diagnosis.
the bedside.16 Thus, 2 years after its publication in 2007, it The Pediatric Risk of Mortality (PRISM) score17 and the
makes sense to validate whether the main TRIPICU Pediatric Logistic Organ Dysfunction (PELOD) score18
recommendation—to consider giving RBC transfusions to were used to describe severity of illness.
stable or stabilized critically ill children without cyanotic Scientific and ethics approval for this study was
heart disease only if their Hb level drops below 70 g/L—is obtained from the institutional review board (ethics com-
currently being applied. mittee) of the Sainte-Justine Hospital Research Center;
The primary aim of study was to determine transfu- due to the observational nature of this study, the Board did
sion practice in a large multidisciplinary PICU (epidemi- not request informed consent.
ology and determinants of RBC transfusion). The
secondary aims were to compare this practice with that
observed 10 years earlier and to assess the degree of appli- Determinants of RBC transfusions
cation of the main TRIPICU recommendation.
Selection of determinants
Before initiating the study, a list of possible determinants
MATERIALS AND METHODS of RBC transfusion was generated; this list was based on a
review of the available literature,2,6,11,19,20 on medical expe-
Study site rience, and on the results of a survey we had previously
The PICU of Sainte-Justine University Hospital is a multi- done.12 In TCs, a variable was considered as a possible
disciplinary 24-bed PICU, serving both medical and sur- determinant only if it was observed before the first RBC
gical specialties. Premature infants are not admitted transfusion in the PICU; determinants could have
directly to the PICU, unless they require cardiac surgery occurred at any time before the transfusion event during
or were discharged home before PICU admission. On the PICU stay. In nTCs, the presence of the same possible
average, there are 1000 admissions per year. There are no determinants was looked for during the entire PICU stay.
transfusion guidelines in our institution, but the principal
investigators of TRIPICU were from Sainte-Justine
Definitions of determinants
Hospital; therefore, physicians in our PICU are well aware
Patient characteristics and the data collected at admission
of the findings from TRIPICU and can decide whether
were considered as potential determinants. In addition,
there is an indication for transfusion based on this
the presence or absence of each of the possible determi-
knowledge.
nants listed below was assessed daily.
The worst Hb was considered to be the lowest Hb
level within the 24 hours before transfusion for TCs and
Study population the lowest Hb level during the entire PICU stay for nTCs.
All consecutive admissions to the PICU, from April 21, Multiple organ dysfunction syndrome (MODS) was
2009, to April 20, 2010, were prospectively screened for defined as the presence of dysfunction of two or more
recruitment. A patient was included unless he or she met organ systems, as defined by Proulx and colleagues.21
one of the exclusion criteria (gestational age less than 40 Hypotension was defined as a systolic blood pressure
weeks at the time of PICU entry, age less than 3 days or below the fifth percentile for age.22,23 Head trauma was
more than 18 years at PICU admission, pregnancy or considered as severe when the Glasgow Coma Score was
admission just after labor). equal to or less than 8. Systemic inflammatory response

366 TRANSFUSION Volume 54, February 2014

 RBC TRANSFUSION IN THE PICU

Multiple logistic regression was

used for the multivariate analysis.
Determinants were included in the
multivariate model if they were clini-
cally relevant, and/or significant in the
univariate analysis, and if they did not
show colinearity with another retained
determinant. A threshold Hb level of
70 g/L was chosen as a determinant in
the multivariate model because it
had the best global value. Interac-
tion terms believed to be clinically rel-
evant were introduced and each was
tested for significance using likelihood
ratio.
All results are two-sided and were
Fig. 1. Patient flow chart. considered significant when the p value
was less than 0.05. All statistical analy-
ses were done by authors PD and TD,
syndrome, sepsis, severe sepsis, and septic shock were using computer software (SPSS statistical software,
defined according to the definitions published by an inter- Release 17.0.1, SPSS, Inc., Chicago, IL).
national pediatric sepsis consensus conference.22 Acute
respiratory distress syndrome was defined according to
RESULTS
the American-European Consensus Conference on acute
respiratory distress syndrome criteria.24 Bacterial and viral A total of 802 patients contributed to 913 consecutive
infections were recorded as per medical notes in the PICU admissions over 1 year (Fig. 1); 71 admissions were
patient chart and by following up all cultures done during excluded (some children met more than one exclusion
PICU stay. criteria). No children were lost to follow-up. A total of 842
admissions (cases) were retained for analysis. At least one
RBC transfusion was given in 144 cases (17.1%).
Justification for RBC transfusions
For the first transfusion event of each TC, the attending
physician was asked a series of questions within 48 hours Determinants of RBC transfusion
after the transfusion to determine why the transfusion was
Univariate analysis
prescribed. The questionnaire included a predefined list
The number of TCs and nTCs with and without possible
of items from which the physician could choose those that
risk factors are reported in Table 1. Significant determi-
justified his or her decision to transfuse and also provided
nants on admission included young age (<12 months),
the possibility to write a justification not mentioned in the
congenital heart disease, severity of illness (highest PRISM
predefined list. When several justifications were selected,
and PELOD scores), and diagnosis of cardiac surgery or
respondents were asked to attribute an order of impor-
shock. Patients with cyanotic heart disease more fre-
tance to them.
quently received transfusions than those with a noncyan-
otic heart disease (39.8% vs. 16.7%; p = 0.001), even
Statistical analysis though there was no significant difference between them
Descriptive statistics were reported as a fraction of the for PRISM and PELOD scores (6.6 ⫾ 5.8 in cyanotic vs.
total population, mean ⫾ standard deviation (SD), and/or 5.7 ⫾ 5.4 in noncyanotic, p = 0.37 and 6.9 ⫾ 6.2 for cyan-
median (range) with interquartile range. Continuous vari- otic vs. 6.3 ⫾ 6.7 for noncyanotic, p = 0.81, respectively).
ables were compared using t test if the variable was nor- After PICU admission, the most significant determi-
mally distributed or Wilcoxon test if not. The chi-square nants for transfusion were a low Hb level, MODS, an
statistic was used for categorical variables. elevated daily PELOD score, presence of at least one organ
For all possible determinants, univariate analysis was dysfunction, and specific conditions (hypotension, severe
done by calculating an odds ratio (OR) with its 95% con- sepsis, septic shock).
fidence interval (95% CI). Dichotomous variables are
presented with a single OR and a p value. Continuous Multivariate analysis
variables (age, PRISM score) were separated into The following determinants were significant in the mul-
categories. tivariate analysis: young age (<12 months), congenital

Volume 54, February 2014 TRANSFUSION 367

DEMARET ET AL.

TABLE 1. Determinants of RBC transfusion: univariate and multivariate analyses

Determinants nTCs TCs Univariate OR (95% CI) p value Multivariate OR (95% CI) p value
Patients’ characteristics
Age (months)*
<12 (D+) 194 71 2.01 (1.26-3.2) 0.003 2.43 (1.19-4.96) 0.015
12-59 (D+) 185 26 0.77 (0.44-1.35) 0.37 1.41 (0.63-3.18) 0.402
60-143 (D+) 143 15 0.58 (0.3-1.11) 0.1 0.39 (0.14-1.14) 0.085
ⱖ144 (reference group: D–) 176 32 Reference Reference
Sex
Male/female 360/338 74/70 0.99 (0.69-1.42) 0.95
Congenital heart disease* (D+/D–) 121/577 50/94 2.54 (1.71-3.77) <0.001 3.61 (1.99-6.55) <0.001
Cyanotic 56/642 37/107 3.96 (2.5-6.3) <0.001
Noncyanotic 65/633 13/131 0.97 (0.52-1.81) 0.915
Data on PICU admission day
PRISM score
>10 (D+) 82 54 7.24 (3.73-14.07) <0.001
6-10 (D+) 181 44 2.67 (1.39-5.15) 0.003
1-5 (D+) 292 33 1.24 (0.63-2.43) 0.53
0 (reference group: D–) 143 13 Reference
PELOD score
>20 (D+) 12 18 23.17 (9.82-54.72) <0.001
11-20 (D+) 117 49 6.47 (93.69-11.35) <0.001
1-10 (D+) 260 57 3.39 (1.98-5.79) <0.001
0 (reference group: D–) 309 20 Reference
Admission diagnosis† (D+/D–)
Respiratory disease 261/436 37/107 0.58 (0.39-0.87) 0.007
Bacterial infection 195/502 42/102 1.06 (0.71-1.57) 0.77
Viral infection 181/514 22/122 0.51 (0.32-0.83) 0.006
Noncardiac surgery 179/516 25/119 0.61 (0.38-0.96) 0.033
Cardiac surgery 74/624 34/110 2.6 (1.66-4.1) <0.001
Seizures 60/638 2/142 0.15 (0.04-0.62) 0.003
Any shock 26/669 23/121 4.89 (2.7-8.85) <0.001
Polytrauma without head trauma 15/681 3/141 0.97 (0.28-3.38) 0.96
Severe head trauma 8/689 3/141 1.83 (0.48-6.99) 0.37
Data during PICU stay‡ (D+/D-)
Worst Hb level ⱕ 70 g/L* (D+/D-) 16/607 60/82 27.76 (15.3-50.5) <0.001 61.3 (27.75-134.7) <0.001
MODS 87/610 75/69 7.6 (5.2-11.3) <0.001
Worst daily PELOD score*
>20 (D+) 18 23 27.37 (11.75-63.8) <0.001 6.33 (1.72-23.25) 0.005
11-20 (D+) 165 59 7.66 (4-14.68) <0.001 3.27 (1.23-8.69) 0.018
1-10 (D+) 253 47 3.98 (2.06-7.68) <0.001 2.76 (1.08-7.04) 0.034
0 (reference group: D-) 257 12 Reference Reference
Respiratory dysfunction* 228/470 61/82 1.52 (1.05-2.19) 0.026 §
Cardiovascular dysfunction* 57/640 57/87 7.36 (4.78-11.31) <0.001 4.66 (2.48-8.76) <0.001
Hematologic dysfunction* 31/666 49/95 11.1 (6.7-18.2) <0.001 4.7 (2.25-9.83) <0.001
Neurologic dysfunction* 85/612 51/93 3.95 (2.62-5.95) <0.001 2.04 (1.09-3.83) 0.026
Gastrointestinal dysfunction* 2/695 6/138 15.11 (3.02-75.64) <0.001 §
Hepatic dysfunction* 24/673 30/114 7.38 (4.2-13.08) <0.001 3.16 (1.4-7.1) 0.005
Renal dysfunction* 7/690 12/132 8.96 (3.46-23.18) <0.001 §
ARDS 29/667 8/135 1.36 (0.61-3.05) 0.449
Hypotension 168/530 70/73 3.03 (2.09-4.38) <0.001
Severe sepsis* 39/658 18/125 2.43 (1.35-4.38) 0.002 §
Septic shock 12/685 11/132 4.76 (2.06-11.01) <0.001
Bacterial infection 208/489 49/94 1.23 (0.84-1.8) 0.296
Viral infection 192/504 25/118 0.56 (0.35-0.88) 0.012
* Determinants selected for the multivariate analysis.
† More than one diagnosis may have been attributed to one patient. Several diagnoses were not included in the analyses (intoxication [30],
arrhythmia [21], intracranial hemorrhage [16], diabetic ketoacidosis [14], brain death [6], etc.).
‡ Data before the first transfusion for TCs; data during the whole PICU stay for nTCs.
§ Determinants not retained in the final model (eliminated one after the other because of a nonsignificant Wald test with an alpha level of
0.05).
ARDS = acute respiratory distress syndrome; D- = total number of cases without determinant; D+ = total number of cases with determinant;
NS = not significant.

heart disease, lowest Hb level of not more than involving the following variables: age, Hb of not more
70 g/L, elevated daily PELOD score, and presence than 70 g/L, and congenital heart disease. Since they
of some organ dysfunctions (cardiovascular, hemato- were not significant, they were excluded from the final
logic, neurologic, or hepatic). We tested the interactions model.

368 TRANSFUSION Volume 54, February 2014

 RBC TRANSFUSION IN THE PICU

77.3 ⫾ 22.2 g/L. The pretransfusion Hb

TABLE 2. Hb as a determinant of RBC transfusion level was more than 70 g/L in 82 TCs
Lowest Hb (g/L) during PICU stay* nTCs TCs Univariate OR (95% CI) p value (57.7%; Fig. 3), among which 46 (56.1%)
All patients had congenital heart disease. The mean
ⱕ70 16 60 165 (46.32-587.73) <0.001
70-80 50 17 14.96 (4.2-53.26) <0.001
pretransfusion Hb was 92.5 ⫾ 14.1 g/L
80-90 79 21 11.7 (3.38-40.48) <0.001 for children with cardiac disease versus
90-100 111 20 7.93 (2.3-27.38) 0.001 69.2 ⫾ 21.5 g/L for children without
100-110 131 18 6.05 (1.74-21.02) 0.005
110-120 104 3 1.27 (0.25-6.42) 0.773
cardiac disease (p < 0.001). The pre-
>120 132 3 Reference transfusion Hb was higher in children
Patients with congenital heart disease with cyanotic heart disease (97 ⫾
ⱕ70 3 3 21 (1.61-273.34) 0.02
70-80 8 7 18.38 (1.94-173.98) 0.011
11.9 g/L) than in children with noncy-
80-90 19 10 11.05 (1.29-94.63) 0.028 anotic heart disease (80 ⫾ 12.4 g/L)
90-100 27 13 10.11 (1.22-83.6) 0.032 (p < 0.001). A total of 110 of the 144 first
100-110 23 14 12.78 (1.55-105.78) 0.018
110-120 14 1 1.5 (0.09-26.01) 0.781
transfusion events (76.4%) were pre-
>120 21 1 Reference scribed during the first 2 days in PICU.
Patients without congenital heart disease
ⱕ70 13 57 243.35 (53.09-1115.5) <0.001
70-80 42 10 13.21 (2.78-62.83) 0.001
Justifications for the first
80-90 60 11 10.18 (2.18-47.42) 0.003 RBC transfusion
90-100 84 7 4.63 (0.94-22.84) 0.06
The questionnaire asking attending
100-110 108 4 2.06 (0.37-11.46) 0.411
110-120 90 2 1.23 (0.17-8.93) 0.836 physicians why they prescribed a RBC
>120 111 2 Reference transfusion was filled out in 138 of the
* Before first transfusion for TCs; during the whole PICU stay for nTCs. 144 TCs (95.8%; Fig. 4). A total of 528 jus-
tifications were given for 138 transfu-
sions (3.8 justifications per transfusion
Hb level as a determinant of RBC transfusion event). A low Hb level was the most frequently cited (112
Data on lowest Hb before the first transfusion (for TCs) or times, 21.2% of all justifications), followed by desire to
during the whole PICU stay (for nTCs) are reported in improve oxygen delivery (79 times, 15%), underlying
Table 2. The unadjusted univariate OR for transfusion disease (69 times, 13.1%), hemodynamic instability (47
increases as the Hb level decreases. The OR for transfusion times, 8.9%), and admission after cardiac surgery (42
of children with a Hb level of not more than 70 g/L was times, 8%). When physicians were asked to choose the
165 (95% CI, 46.32-587.73) compared with the reference most important justification, a low Hb level (55/138,
group (i.e., children with a lowest Hb level >120 g/L). This 39.9%), desire to improve oxygen delivery (23/138, 16.7%),
strong association was more marked in children without and hemodynamic instability (20/138, 14.5%) were given
than those with congenital heart disease (OR, 243.35 the highest hierarchy.
[95% CI, 53.09-1115.5] vs. 21 [95% CI, 1.61-273.34],
respectively).
Compliance to TRIPICU recommendations
The receiver operating characteristic (ROC) curve
(Fig. 2) illustrates the capacity of the lowest Hb to predict A comparison of our population with that of TRIPICU is
RBC transfusion. The area under the curve (AUC) is 0.827 reported in Table 4. We examined the first transfusion of
(95% CI, 0.79-0.87), which is significantly different each TC (Fig. 4). Five transfusions (3.6%) were prescribed
from the no-discrimination line (p < 0.001). The thresh- in stable patients with a Hb level more than 70 g/L. For
old with the best global predictive value of RBC transfu- these transfusions, the principal justifications were hypo-
sion was determined using the Youden index, whose volemia (pretransfusion Hb level of 74 g/L), marrow trans-
value of 0.475 corresponded with a Hb level of 93.5 g/L plant (78 g/L), before surgery with a high risk of
(sensitivity, 74.6%; specificity, 72.9%). When considering hemorrhage (73 g/L), surgeon’s request (88 g/L), or cere-
only patients without congenital heart disease, the bral edema (80 g/L). Seventy-five transfusions (52.1%)
ROC curve improves, with an AUC of 0.896; the Youden were given to patients with a Hb level of more than 70 g/L
index is then 0.252, corresponding to a Hb level of in whom TRIPICU recommendations were not applicable
86.5 g/L. (according to the prescribing physician) because of hemo-
dynamic instability (53 cases, 38.4%), active bleeding (18,
13%), cyanotic congenital heart disease (13, 9.4%), con-
Data on RBC transfusions genital hemolytic anemia (four, 2.9%), extracorporeal
Data on transfusions are reported in Table 3. A total of 578 membrane oxygenation (three, 2.2%), or coronaropathy
transfusions were given during the 1-year study period. (congenital atresia of the left main coronary artery; one,
The mean Hb level before the first transfusion was 0.7%).

Volume 54, February 2014 TRANSFUSION 369

DEMARET ET AL.

since the publication of the TRIPICU

study.14 At least one RBC transfusion
was given to 17% of our patients.
The mean pretransfusion Hb was
77.3 ⫾ 22.2 g/L. We identified several
determinants of RBC transfusion: young
age, congenital heart disease, Hb level of
not more than 70 g/L, high daily PELOD
score, and some organ dysfunctions
were independently associated with a
higher OR for RBC transfusion. The
three most important justifications for
RBC transfusion were a low Hb level,
insufficient oxygen delivery, and hemo-
dynamic instability. A small number
of transfusions were administered to
stable patients with a Hb level greater
than 70 g/L.

Knowledge transfer
Knowledge transfer from clinical
research findings to the bedside can
Fig. 2. Lowest Hb level as a predictor of RBC transfusion. ROC curve using the lowest involve substantial delay. The Transfu-
Hb level (before the first transfusion for TCs, during the whole PICU stay for nTCs) sion Requirements in Critical Care
to differentiate between transfused and nontransfused children. AUC = 0.827 (95% (TRICC) study, published in 1999,
CI, 0.79-0.87; p < 0.001). reported that maintaining Hb levels
between 70 and 90 g/L was safe in most
critically ill adults.25 Pretransfusion Hb levels were
TABLE 3. Data on RBC transfusions in patients reported to be 84 ⫾ 13 g/L by Vincent and colleagues in
transfused in PICU
19993 and 86 ⫾ 17 g/L by Corwin and colleagues in 2000
Number of RBC transfusions 578
Number (%) of patients who 144 (17.1) to 2001.2 These values were not different from the
received at least one 86 ⫾ 13 g/L observed by Hebert and coworkers before the
transfusion
TRICC trial.25 Meaningful changes in transfusion practices
Incidence density of 11.2
transfusion (transfusion may not have been captured, the time period between
events/100 patient-days)* publication of TRICC and these two descriptive studies
Volume of blood during first 12.5 ⫾ 8.9; 11.1 (7.3-15)
being too short. Netzer and coworkers26 studied transfu-
transfusion (mL/kg)†
Lowest Hb level within 24 hr 77.3 ⫾ 22.2; 77.5 (63.8-94) sion practices during a 10-year period (1997 to 2007). They
before first transfusion (g/L)† found that the period of large-magnitude change
Time from PICU admission to 1.3 ⫾ 2.1; 1 (0-1)
occurred in the first 5 years after publication of TRICC and
first transfusion (days)†
PRISM score within 24 hr 8 ⫾ 6.3; 7 (3-12) that subsequent smaller magnitude change persisted for
before first transfusion† another 3 years.
PELOD score within 24 hr 8.6 ⫾ 8.4; 10 (1-12)
Our study is the first to address the question of
before first transfusion†
Length of storage of RBC 15.7 ⫾ 9.4; 13 (9-22) knowledge application of TRIPICU findings. The propor-
units (days)†‡ tion of children transfused was 17.1%, which is quite
* Calculated using the total number of transfusion events (nu- similar to the 14% we observed 10 years earlier
merator) and the sum of the days that each child included in (p = 0.078).11 Severity of illness as reflected by the admis-
the study spent in PICU (denominator).
† Mean ⫾ SD; median (interquartile range). sion PRISM score (6 ⫾ 5.8 vs. 5.7 ⫾ 6.3 in 2000), as well
‡ Data available for 354 transfusion events. as mean age at admission (71.9 ⫾ 72 months vs.
73.2 ⫾ 68.4 months) and mean Hb level at PICU entry
(111.5 ⫾ 24.7 g/L vs. 114 ⫾ 25 g/L) were comparable at
these two time points. However, the mean Hb level
DISCUSSION
before first transfusion was 77.7 ⫾ 22.2 g/L in 2010, a
This prospective observational study is the first to evalu- value markedly lower than the 88 ⫾ 26 g/L observed
ate transfusion practice in a large multidisciplinary PICU 10 years earlier.11 It thus seems that transfusion practices

370 TRANSFUSION Volume 54, February 2014

 RBC TRANSFUSION IN THE PICU

Fig. 3. Hb level within 24 hours before first RBC transfusion. Congenital heart disease: children with cyanotic or noncyanotic heart
disease. (䊏) Congenital heart disease; ( ) no congenital heart disease.

144 TCs
6 missing questionnaires (4.2%)
- 3 with lowest pretransfusion Hb < 70g/L
- 2 with a cyanotic congenital heart disease
- 1 for which TRIPICU recommendations could
First transfusion apparently be applied
justified by physician in
138 cases (95.8%)

Pretransfusion Hb Pretransfusion Hb
70g/L: n=58 (42%) > 70g/L: n=80 (58%)
(Congenital heart disease: n=45,
including 30 cases of cardiac surgery)

Non adherence to TRIPICU Transfusions for which TRIPICU

recommendations: recommendations were not applicable:
- Hypovolemia (1) (Hb 74g/L) - Hemodynamic instability (53, 38.4%)
- Bone marrow transplant (1) (Hb 78g/L) - Active bleeding (18, 13%)
- Preparation for a surgery with a high - Cyanotic congenital heart disease (13, 9.4%)
hemorrhagic risk (1) (Hb 73g/L) - Congenital hemolytic anemia (4, 2.9%)
- Surgeon request (1) (Hb 88g/L) - ECMO (3, 2.2%)
- Cerebral edema (1) (Hb 80g/L) - Coronaropathy* (1, 0.7%)

Fig. 4. Justifications for first RBC transfusions (according to prescribing physician). ECMO = extracorporeal membrane oxygen-
ation. *Congenital atresia of the left main coronary artery.

in our PICU became more conservative over time. ture describing the hazards of transfusion). However,
Factors other than TRIPICU may have influenced our TRIPICU remains the only RCT comparing two transfu-
transfusion practice during the past 10 years (like sion strategies in PICU. We therefore believe that
changes in physicians on staff, inferences from the adult TRIPICU is the main factor that has led to changes in
literature [such as the TRICC study] or published litera- transfusion practices in our unit.

Volume 54, February 2014 TRANSFUSION 371

DEMARET ET AL.

TABLE 4. Comparison of some characteristics of the patients enrolled in the TRIPICU study with the patients
included in this study*
TRIPICU study: RBC transfusion strategy
Clinical data Restrictive (n = 320) Liberal (n = 317) This study (n = 842)
PICU admission age (month) 35.8 ⫾ 46.2 39.6 ⫾ 51.9 71.9 ⫾ 72
Male sex 190 (59) 191 (60) 434 (51.5)
Admission PRISM score 9.4 ⫾ 6.7 9.1 ⫾ 6.7 6 ⫾ 5.8
Congenital cyanotic heart disease† 0 (0) 0 (0) 93 (11)
Admission after cardiac surgery 63 (20) 62 (20) 108 (12.8)
Admission after noncardiac surgery 60 (19) 64 (20) 204 (24.2)
Admission for medical reason 197 (61) 191 (60) 530 (62.9)
Baseline Hb (g/L)‡ 80 ⫾ 10 80 ⫾ 9 111.5 ⫾ 24.7
Lowest Hb > 95 g/L§ 0 (0) 0 (0) 455 (54)
PICU length of stay (day) 9.5 ⫾ 7.4 9.9 ⫾ 7.9 6.1 ⫾ 14.5
Survivors with PICU stay ⱕ1 day|| 0 (0) 0 (0) 68 (8.1)
* Data are reported as number (%) or mean ⫾ SD.
† Patients with cyanotic heart disease were excluded from TRIPICU.
‡ At randomization for the patients in TRIPICU, at PICU admission for TCs in this study.
§ Before randomization for the patients in TRIPICU, before the first transfusion for TCs in this study, during the entire PICU stay for nTCs in
this study (patients were considered for inclusion in the TRIPICU study only if their Hb level was < 95/L within the first 7 days in PICU).
|| Patients were excluded from TRIPICU if they were expected to stay less than 24 hours in PICU.

Main justifications for RBC transfusions The concept of goal-directed RBC therapy is an
According to attending physicians, a low Hb level was attractive approach. It suggests that RBCs should be trans-
the main justification of 40% of the first transfusion fused to attain a given “physiologic” goal and not only to
events. Bateman and coworkers4 reported the same reach a given Hb level.32 Different markers of adequate
finding in 2005. Observed practice patterns can differ oxygen delivery have been suggested as goals. In our
from stated practice patterns, but this was not the case in study, the desire to improve oxygen delivery was cited 79
our experience: we also observed at the bedside that the times as a justification for the first transfusion event (79/
Hb level was the strongest determinant of RBC transfu- 138, 57.2%), which implies that there was no stated intent
sion (OR if Hb ⱕ 70 g/L, 61.3; 95% CI, 27.75-134.66; to improve oxygen delivery in 42.8% of the first transfu-
p < 0.001). Improvement of oxygen delivery and concern sion events. Among the 112 first transfusion events justi-
about hemodynamic instability were two other fre- fied by a low Hb level, only 64 (57.1%) were also justified by
quently evoked justifications for physicians to prescribe the desire to improve oxygen delivery. Among the 47 trans-
transfusion. fusions justified by hemodynamic instability, only 31
The fact that RBC transfusion increases blood oxygen (66%) were also justified by the desire to improve oxygen
content is indisputable. However, this increase does not delivery. This suggests that goal-directed transfusion is
necessarily improve oxygen delivery to cells. Storage not a concept applied in practice. Future studies are
lesion (i.e., modifications of RBCs during storage) can required to promote the use of this concept and to deter-
contribute to this apparent contradiction.27,28 For mine appropriate goals of transfusion.
example, adverse physiologic modifications of intra-RBC
Hb can cause compromised NO signaling28,29 and lead to RBC transfusion in children with congenital
acute microvascular dysregulation. Thus, it is not neces- heart disease
sarily true that those intending to increase O2 consump- There is no solid evidence with respect to the threshold Hb
tion will attain this goal with RBC transfusion. that should prompt RBC transfusion in children with con-
Many intensivists believe that they should maintain a genital heart disease. Our study showed that transfusion
higher Hb level in hemodynamically unstable patients. strategies differ between children with and without con-
Few hard data support this point of view. Two trials, one in genital heart disease (Fig. 3). This difference is largely
children30 and another in adults31 with severe sepsis or attributable to the presence of cyanotic heart disease in 37
septic shock, have shown benefit of RBC transfusion to of 50 transfused children with cardiac disease (74%). We
attain a hematocrit level of more than 30% if the central also observed a significantly different pretransfusion Hb
venous O2 saturation remained below 70% after initial level between children with cyanotic or noncyanotic heart
resuscitation. However, the exact role of RBC transfusion disease. Such difference was also found in a recent
in these trials is unclear. We presently do not know what scenario-based survey.33
Hb concentration should prompt intensivists to prescribe While our data suggest that a higher threshold Hb is
an RBC transfusion in unstable children. preferred for children with congenital heart disease, there

372 TRANSFUSION Volume 54, February 2014

 RBC TRANSFUSION IN THE PICU

is little evidence this is appropriate. A subgroup analysis of of the TRIPICU study (Table 4). This could be a limitation
TRIPICU involving 125 children suggested that a Hb level to the assessment we made of the application of the main
of 70 g/L is safe for stable critically ill children with non- TRIPICU recommendation. However, in our analysis, only
cyanotic heart disease.34 In another RCT including 60 chil- cases fulfilling the inclusion criteria of the TRIPICU study
dren with univentricular physiology randomized after were considered as transfused while outside TRIPICU rec-
cardiac surgery to a liberal (transfusion threshold of ommendations (i.e., with a Hb level >70 g/L; Fig. 4).
130 g/L) or a restrictive group (90 g/L), O2 extraction rate Our study also has several strengths. This is the first
was slightly higher in the restrictive group, but median study evaluating transfusion practices in PICU since
and peak blood lactate were almost similar in both TRIPICU was published. The study included all consecu-
groups.35 At present, the only possible conclusion is that it tive PICU admissions over a 1-year period, which resulted
is probably safe to use a transfusion threshold of 90 g/L in in a case mix with a limited risk of selection bias and no
children with cyanotic heart malformations and 70 g/L in influence due to seasonal variation. The list of possible
those with noncyanotic malformations who are stable; determinants of RBC transfusion was decided upon
better evidence is necessary to determine ideal transfu- before the study was initiated. Rigorous attention was
sion practice in children with congenital heart disease. paid to temporal relationship between all determinants
and RBC transfusion because a variable has the potential
to be a determinant only if it occurs before the transfusion
Strengths and limitations of our study event. Finally, the prospective nature of our study was a
Our study has several limitations. It was single center, major asset to minimize information bias.
which limits its external validity; however, our critical care In conclusion, this study describes transfusion prac-
unit is comparable to most multidisciplinary university- tices in a representative academic PICU at a time point 2.5
affiliated North American PICUs with regard to case mix years after the TRIPICU study was published. We found
and severity of illness. Second, the principal investigator that young age, congenital heart disease, a Hb level of not
of TRIPICU is on staff in our PICU, which may have mark- more than 70 g/L, a high daily PELOD score, and organ
edly influenced local transfusion practice and sped up dysfunctions were independently associated with RBC
knowledge transfer. Third, patients readmitted to PICU transfusion in PICU. The proportion of patients receiving
more than 24 hours after discharge were considered as at least one RBC transfusion in our PICU did not signifi-
new cases. It can be argued that this approach leads to cantly change over a 10-year period, but the pretransfu-
biased results because of correlation between cases. sion Hb decreased significantly. Transfusion practices
However, we undertook a sensitivity model using a gener- differed in children with and without congenital heart
alized estimating equation approach36 and showed that disease and also between those with and without cyanotic
considering these patients as one case would have malformations. Finally, we observed that the majority of
resulted in findings similar to those of our multivariate the first transfusion events were prescribed according to
model. Fourth, variables occurring before PICU admis- recent recommendations. A multicenter study would be
sion may have been determinants we did not consider. required to better evaluate transfusion practices in PICU
Our study was designed to find determinants of RBC and to ascertain general compliance to evidence-based
transfusion in PICU and purposefully did not take into recommendations.
account pre-PICU variables because intensivists can
change only what happens in PICU. Fifth, the number of ACKNOWLEDGMENTS
TCs with congenital heart disease is relatively small,
We thank Nicole Poitras and Mariana Dumitrascu for their
making it difficult to distinguish the effects of cyanotic and
support in the realization of this study.
noncyanotic heart disease on transfusion practices. Sixth,
memory bias may have occurred in the part focusing on
justifications for RBC transfusion given by prescribing CONFLICT OF INTEREST
physicians. Even if the questionnaire was administered None.
within 2 days after transfusion, physicians sometimes
waited days before filling it in. Moreover, when physicians
REFERENCES
were completing the questionnaire, being asked to
provide justifications may have motivated them to give 1. Carson JL, Grossman BJ, Kleinman S, Tinmouth AT,
more justifications than they actually had at the time Marques MB, Fung MK, Holcomb JB, Illoh O, Kaplan LJ,
of transfusion. Seventh, hemodynamic instability was Katz LM, Rao SV, Roback JD, Shander A, Tobian AA, Wein-
assessed by physicians using their own criteria (not nec- stein R, Swinton McLaughlin LG, Djulbegovic B; Clinical
essarily matching TRIPICU’s criteria of stability); this may Transfusion Medicine Committee of the AABB. Red blood
have led to overestimation of compliance to TRIPICU rec- cell transfusion: a clinical practice guideline from the
ommendations. Finally, our case mix is different than that AABB. Ann Intern Med 2012;157:49-58.

Volume 54, February 2014 TRANSFUSION 373

DEMARET ET AL.

2. Corwin HL, Gettinger A, Pearl RG, Fink MP, Levy MM, Transfusion strategies for patients in pediatric intensive
Abraham E, MacIntyre NR, Shabot MM, Duh MS, Shapiro care units. N Engl J Med 2007;356:1609-19.
MJ. The CRIT Study: anemia and blood transfusion in the 15. Grol R, Grimshaw J. From best evidence to best practice:
critically ill—current clinical practice in the United States. effective implementation of change in patients’ care.
Crit Care Med 2004;32:39-52. Lancet 2003;362:1225-30.
3. Vincent JL, Baron JF, Reinhart K, Gattinoni L, Thijs L, Webb 16. Esteban A, Ferguson ND, Meade MO, Frutos-Vivar F, Apez-
A, Meier-Hellmann A, Nollet G, Peres-Bota D; ABC (Anemia teguia C, Brochard L, Raymondos K, Nin N, Hurtado J,
and Blood Transfusion in Critical Care) Investigators. Tomicic V, González M, Elizalde J, Nightingale P, Abroug F,
Anemia and blood transfusion in critically ill patients. Pelosi P, Arabi Y, Moreno R, Jibaja M, D’Empaire G, Sandi
JAMA 2002;288:1499-507. F, Matamis D, Montañez AM, Anzueto A; VENTILA Group.
4. Bateman ST, Lacroix J, Boven K, Forbes P, Barton R, Evolution of mechanical ventilation in response to clinical
Thomas NJ, Jacobs B, Markovitz B, Goldstein B, Hanson research. Am J Respir Crit Care Med 2008;177:170-7.
JH, Li HA, Randolph AG; Pediatric Acute Lung Injury and 17. Pollack MM, Ruttimann UE, Getson PR. Pediatric risk of
Sepsis Investigators Network. Anemia, blood loss, and mortality (PRISM) score. Crit Care Med 1988;16:1110-6.
blood transfusions in North American children in the 18. Leteurtre S, Martinot A, Duhamel A, Proulx F, Grandbas-
intensive care unit. Am J Respir Crit Care Med 2008;178:26- tien B, Cotting J, Gottesman R, Joffe A, Pfenninger J,
33. Hubert P, Lacroix J, Leclerc F. Validation of the paediatric
5. Kneyber MC, Hersi MI, Twisk JW, Markhorst DG, Plotz FB. logistic organ dysfunction (PELOD) score: prospective,
Red blood cell transfusion in critically ill children is inde- observational, multicentre study. Lancet 2003;362:
pendently associated with increased mortality. Intensive 192-7.
Care Med 2007;33:1414-22. 19. Hebert PC, Wells G, Martin C, Tweeddale M, Marshall J,
6. Goodman AM, Pollack MM, Patel KM, Luban NL. Pediatric Blajchman M, Pagliarello G, Schweitzer I, Calder L. A Cana-
red blood cell transfusions increase resource use. J Pediatr dian survey of transfusion practices in critically ill patients.
2003;142:123-7. Transfusion Requirements in Critical Care Investigators
7. White M, Barron J, Gornbein J, Lin JA. Are red blood cell and the Canadian Critical Care Trials Group. Crit Care Med
transfusions associated with nosocomial infections in 1998;26:482-7.
pediatric intensive care units? Pediatr Crit Care Med 2010; 20. Rao MP, Boralessa H, Morgan C, Soni N, Goldhill DR, Brett
11:464-8. SJ, Boralessa H, Contreras M; North Thames Blood Interest
8. Vamvakas EC, Blajchman MA. Transfusion-related immu- Group. Blood component use in critically ill patients. Ana-
nomodulation (TRIM): an update. Blood Rev 2007;21:327- esthesia 2002;57:530-4.
48. 21. Proulx F, Fayon M, Farrell CA, Lacroix J, Gauthier M. Epi-
9. Shaz BH, Stowell SR, Hillyer CD. Transfusion-related acute demiology of sepsis and multiple organ dysfunction syn-
lung injury: from bedside to bench and back. Blood 2011; drome in children. Chest 1996;109:1033-7.
117:1463-71. 22. Goldstein B, Giroir B, Randolph A. International pediatric
10. Li G, Rachmale S, Kojicic M, Shahjehan K, Malinchoc M, sepsis consensus conference: definitions for sepsis and
Kor DJ, Gajic O. Incidence and transfusion risk factors for organ dysfunction in pediatrics. Pediatr Crit Care Med
transfusion-associated circulatory overload among medical 2005;6:2-8.
intensive care unit patients. Transfusion 2011;51:338-43. 23. Goldstein B, Giroir B, Randolph A. Values for systolic blood
11. Armano R, Gauvin F, Ducruet T, Lacroix J. Determinants of pressure. Pediatr Crit Care Med 2005;6:500-1.
red blood cell transfusions in a pediatric critical care unit: 24. Bernard GR, Artigas A, Brigham KL, Carlet J, Falke K,
a prospective, descriptive epidemiological study. Crit Care Hudson L, Lamy M, LeGall JR, Morris A, Spragg R. The
Med 2005;33:2637-44. American-European Consensus Conference on ARDS.
12. Laverdiere C, Gauvin F, Hebert PC, Infante-Rivard C, Definitions, mechanisms, relevant outcomes, and clinical
Hume H, Toledano BJ, Guertin MC, Lacroix J; Canadian trial coordination. Am J Respir Crit Care Med 1994;
Critical Care Trials Group. Survey on transfusion practices 149(Pt 1):818-24.
of pediatric intensivists. Pediatr Crit Care Med 2002;3:335- 25. Hebert PC, Wells G, Blajchman MA, Marshall J, Martin C,
40. Pagliarello G, Tweeddale M, Schweitzer I, Yetisir E. A multi-
13. Nahum E, Ben-Ari J, Schonfeld T. Blood transfusion policy center, randomized, controlled clinical trial of transfusion
among European pediatric intensive care physicians. requirements in critical care. Transfusion Requirements in
J Intensive Care Med 2004;19:38-43. Critical Care Investigators, Canadian Critical Care Trials
14. Lacroix J, Hebert PC, Hutchison JS, Hume HA, Tucci M, Group. N Engl J Med 1999;340:409-17.
Ducruet T, Gauvin F, Collet JP, Toledano BJ, Robillard P, 26. Netzer G, Liu X, Harris AD, Edelman BB, Hess JR, Shan-
Joffe A, Biarent D, Meert K, Peters MJ; TRIPICU holtz C, Murphy DJ, Terrin ML. Transfusion practice in the
Investigators; Canadian Critical Care Trials Group; Pediat- intensive care unit: a 10-year analysis. Transfusion 2010;50:
ric Acute Lung Injury and Sepsis Investigators Network. 2125-34.

374 TRANSFUSION Volume 54, February 2014

 RBC TRANSFUSION IN THE PICU

27. van de Watering L. Red cell storage and prognosis. Vox 32. Lacroix J, Demaret P, Tucci M. Red blood cell transfusion:
Sang 2011;100:36-45. decision making in pediatric intensive care units. Semin
28. Bennett-Guerrero E, Veldman TH, Doctor A, Telen MJ, Perinatol 2012;36:225-31.
Ortel TL, Reid TS, Mulherin MA, Zhu H, Buck RD, Califf 33. Harrington K, Farrell CA, Poirier N, Ducruet T, Lacroix J.
RM, McMahon TJ. Evolution of adverse changes in stored Survey on red-cell transfusion practices after paediatric
RBCs. Proc Natl Acad Sci U S A 2007;104:17063-8. cardiac surgery. Pediatr Crit Care Med 2011;12:A82.
29. Weinberg JA, Barnum SR, Patel RP. Red blood cell age and 34. Willems A, Harrington K, Lacroix J, Biarent D, Joffe AR,
potentiation of transfusion-related pathology in trauma Wensley D, Ducruet T, Hébert PC, Tucci M; TRIPICU
patients. Transfusion 2011;51:867-73. investigators; Canadian Critical Care Trials Group; Pediat-
30. de Oliveira CF, de Oliveira DS, Gottschald AF, Moura JD, ric Acute Lung Injury and Sepsis Investigators (PALISI)
Costa GA, Ventura AC, Fernandes JC, Vaz FA, Carcillo JA, Network. Comparison of two red-cell transfusion strategies
Rivers EP, Troster EJ. ACCM/PALS haemodynamic support after pediatric cardiac surgery: a subgroup analysis. Crit
guidelines for paediatric septic shock: an outcomes com- Care Med 2010;38:649-56.
parison with and without monitoring central venous 35. Cholette JM, Rubenstein JS, Alfieris GM, Powers KS, Eaton
oxygen saturation. Intensive Care Med 2008;34: M, Lerner NB. Children with single-ventricle physiology do
1065-75. not benefit from higher hemoglobin levels post cavopul-
31. Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, monary connection: results of a prospective, randomized,
Knoblich B, Peterson E, Tomlanovich M; Early Goal- controlled trial of a restrictive versus liberal red-cell trans-
Directed Therapy Collaborative Group. Early goal-directed fusion strategy. Pediatr Crit Care Med 2011;12:39-45.
therapy in the treatment of severe sepsis and septic shock. 36. Zeger SL, Liang KY. Longitudinal data analysis for discrete
N Engl J Med 2001;345:1368-77. and continuous outcomes. Biometrics 1986;42:121-30.

Volume 54, February 2014 TRANSFUSION 375