TRANSFUSION PRACTICES

Calculating the required transfusion volume in children

Patrick Davies, Simon Robertson, Shilpa Hegde, Rosemary Greenwood, Edwin Massey, and
Peter Davis

BACKGROUND: The traditional method of calculating

blood volume for pediatric transfusion in the UK is
weight (kg) aimed increment in hemoglobin concentration (Hb; g/dL) the transfusion factor, usually quoted
at 3 or 4. This equation is without evidence base. The
aim was to assess how the volume of red cells (RBCs)
affects the increase in serum Hb in children and to
devise a formula that allows accurate volume
calculation.
STUDY DESIGN AND METHODS: All pediatric intensive care charts for 2 years were examined
retrospectively. The immediate pre- and posttransfusion
Hb estimations and the precise volumes of RBC transfused were recorded. Fluid boluses and hemorrhagic
loss during the transfusion were documented.
RESULTS: A total of 7679 patient charts were examined with a total of 564 transfusions. All patients who
were bleeding, had drain losses, or had concurrent
colloid infusions were excluded, giving 379 data points.
The correlation gradient between mL per kg blood
transfused and increase in Hb was 5.02. There was no
significant association between effect and patient
weight, age, starting Hb, transfusion time, or sex. No
significant difference was found in Hb at 1 and 7 hours
posttransfusion.
CONCLUSIONS: The following equation should be
used to calculate transfusion volumes: weight
(kg) increment in Hb (g/dL) 3/(hematocrit [Hct] level
of RBCs). This predicts that with a UK standard Hct of
0.6, 10 mL/kg gives an increment of 2 g/dL. Care must
be taken not to risk hypervolemia, while minimizing
donor exposure. Hb estimation 1 hour after transfusion
is the same as 7 hours after transfusion.

212 TRANSFUSION

Volume 47, February 2007

n contrast to adults, children who need blood transfusions receive a calculated volume of blood
depending on their weight, instead of a whole
number of units. This makes the volume calculation
formula crucial, because incorrect calculation risks circulatory overload or multiple transfusions with additional
cost, resource use, and exposure implications. Blood
transfusions have inherent risks that need to be minimized as far as possible. The aim must be to optimize the
outcome of the transfusions whilst minimizing the
number of donors each patient is exposed to.
The calculation of transfusion volumes in children
has not previously been evidence-based. A search on
Medline, on OVID, through textbooks, and of personal
communication from oncologists and hematologists performed before starting this study revealed no studies on
which practice could be based. In the UK, many pediatric
unit protocols use a calculation of (weight of the patient in
kg) (difference in hemoglobin [Hb] to be achieved in
g/dL) a transfusion factor, usually either 3 or 4. Some
individual pediatricians use 20 mL per kg for all patients.
The British Committee for Standards in Haematology1
guidelines advise a transfusion factor of 3 for children, and
Forfar and Arneils Textbook of Paediatrics2 advises a factor
of 4. There is even interdepartmental factor variation
within individual hospitals. In a recent study giving
written scenarios to 134 European pediatric intensive care

ABBREVIATION: PICU = pediatric intensive care unit.

From the Paediatric Intensive Care Unit, Bristol Royal Hospital
for Children, the Research and Development Support Unit,
United Bristol Healthcare Trust, and the National Blood Service
and Bristol Royal Infirmary, Bristol, United Kingdom.
Address reprint requests to: Dr Patrick Davies, e-mail:
[email protected].
This study was funded by the United Bristol Healthcare
Trust Research and Development Office.
Received for publication February 6, 2006; revision
received June 29, 2006, and accepted June 29, 2006.
doi: 10.1111/j.1537-2995.2007.01091.x
TRANSFUSION 2007;47:212-216.

TRANSFUSION IN CHILDREN

consultants, all transfusions requested were for between

10 and 20 mL per kg.3
Children on the pediatric intensive care unit (PICU)
routinely need transfusions, occasionally due to active
blood loss, but more commonly to optimize all possible
parameters in a critically ill child. We noted on the PICU
in Bristol, UK, that with use of our local formula of
weight difference 4, we regularly did not achieve our
desired increase in Hb. The intensive care setting requires
frequent blood tests, including blood gases, enabling realtime tracking of the effect of giving a transfusion. We
decided to analyze the increase in Hb that was achieved
with each transfusion given and to investigate whether the
conventional formula was accurate, and if not, whether
we could devise a simple and more accurate calculation
method.

MATERIALS AND METHODS

All observation charts on the PICU at the Bristol Royal
Hospital for Children for all patients over a 24-month
period from May 2002 onward were examined retrospectively. The intensive care records detail all fluid
input and output by type and volume, as well as observations and blood gas measurements. We took volumes from
the nursing charts that recorded the exact volumes given,
rather than the prescriptions or the calculations on which
they were based. The red blood cell (RBC) specification
did not change during this period. They were resuspended
in optimal additive solution (SAG-M) to a hematocrit (Hct)
level of 0.6 (range, 0.55-0.65).
Patient demographics included age, weight, and sex.
Variables documented for each transfusion episode
included volume of RBCs given, number of hours over
which the transfusion was given, and the immediate pretransfusion, posttransfusion, and 6-hour posttransfusion
blood gas estimations of serum Hb. Blood gas sampling
from an arterial line requires the collection of only 0.5 mL
of blood by syringe, discarding blood and saline in the
line. All colloid infusions and drain losses for the transfusion period, the 5 hours after, and the whole day involved
were documented.
We excluded from analysis all patients for whom there
were confounding variables associated with accelerated
RBC destruction, RBC loss, or changes in blood volume
that would be expected to affect the post transfusion Hb.
Criteria for exclusion were
1.

2.
3.

Patients who received any colloid infusions (4.5%

human albumin solution), platelets, fresh-frozen
plasma, or cryoprecipitate) or plasma exchange
during the transfusion period.
Patients with active bleeding.
All patients on renal replacement therapy or extracorporeal membrane oxygenation.

4.
5.

All patients with a diagnosis of active hemolysis.

Patients in whom there had been any drain losses.
This included chest drains and postoperative drains;
for the purposes of the study these could not be guaranteed to be blood-free losses and were therefore
excluded. Crystalloid fluid boluses are not given on
our unit. Enteral feeds and intravenous maintenance
fluids were not corrected for in the analysis, because
all patients require and receive some form of hydration maintenance. Urine output was also not
analyzed.

Hb measurements were performed on a blood gas

machine (ABL 700, Radiometer A/S, Brnshj, Denmark),
from which clinicians on the Bristol PICU routinely base
patient transfusion requirements. Hb measurements from
the blood gas machine have a daily in-house calibration
and a monthly external calibration. The calibrations for
the study period were found to be always within the laboratory limits. Local research ethics committee approval
and in-house research and development committee
approval were granted. Data were inputted to computer
spreadsheet software (Microsoft Excel 2003, Microsoft
Corp., Redmond, WA) and analyzed with computer software (STATA, Version 7, StataCorp LP).

Statistical analysis
Associations between variables were calculated with Pearsons correlation coefficient and examined with scatter
plots. To quantify the association between volume of RBC
transfused and incremental increase in Hb, a linear regression model was used in which the increment was modeled
as the dependent variable, and volume of transfusion
divided by patient weight was used as the independent
variable. The constant term was excluded from the regression model to allow for no change to be predicted from no
transfusion.
Some patients had repeated blood transfusions
within this cohort. To include all the data, but at the same
time deal with the assumption that separate data points
for the same individual might not be independent, robust
standard errors were calculated with STATA Version 7.

RESULTS
A total of 7679 charts were examined from 1494 admissions, on which 564 blood transfusions were identified
(transfusion rate, 37.8%). A total of 185 transfusions were
excluded due to the above exclusion criteria (Fig. 1). A
total of 94,476 mL of blood was transfused (mean,
167 mL). Demographic data for the patients receiving
analyzed transfusion episodes are shown in Table 1.
For each of the 379 valid data sets the transfusion
factor was calculated by dividing the mL per kg of blood
Volume 47, February 2007

TRANSFUSION 213

DAVIES ET AL.

transfused by the increment achieved in the individual

patient. Increment and mL per kg were strongly associated with each other with a Pearson correlation coefficient
of 0.64. A scatter plot was drawn and a regression line
fitted. The gradient of the regression model was 0.1991
with p < 0.001 and robust standard error 0.0054, which
equates to a transfusion factor of 5.02 (95% confidence
interval, 4.77-5.30; Fig. 2).
There was a significant association between the age in
days of the patient and the transfusion factor (p = 0.049);

Fig. 1. Data flowchart.

however, this has a slope of 1.2 10-4 (transfusion

factor = 5.252 - (age in days 1.2 10-4)). This implies a
change in transfusion factor of -0.04 per year, which is not
clinically significant. This is demonstrated in Fig. 3.
There was no association between the transfusion
factor and other external variables, including weight,
length of transfusion, and starting Hb. There was no difference between the effect of transfusions given to male or
to female patients. Analysis of increase in Hb measured as
a percentage of the starting Hb found a correlation coefficient of 0.63 (SE, 17.6% of starting Hb), which was a less
accurate correlation than by simple increment in g per
dL Hb.
Analysis by type of patient showed no significant difference in transfusion factor by patient category. Cardiac
patients (mostly postoperative; n = 185; mean factor, 4.93),
sepsis patients (n = 60; mean, 5.35), and other patients
(n = 134; mean, 5.04) had no significant difference.
Transfusions were given over a mean duration of
3.4 hours (median, 3 hr) with a standard deviation (SD) of
1.5 hours. Patients who had multiple transfusions had a
mean intertransfusion time of 8.9 days (median, 3 days),
and statistical analysis showed that each transfusion
episode was independent.
We had data of Hb estimation a mean of 2 hours
22 minutes (median, 2 hr) before each transfusion, and
1 hour 23 minutes (median, 1 hr) after transfusion. Analysis of Hb concentrations taken a mean of 7 hours 47
minutes (median, 6 hr) after the end of the transfusion
was no different from those 1 hour after (difference in
g/dL, mean 0.2, median 0.1, SD 0.7; Fig. 4).

DISCUSSION
In this consecutive, single-center study we have described
how transfusions affect Hb concentration in children,
across a broad age and size range. This should help pediatricians and hematologists to practice evidence-based
medicine in this important and common field of clinical
care.
Published data on which to assess the relative risks
of undertransfusion versus those of overtransfusion are
Fig. 2. Scatter plot of transfusion volume (mL/kg) versus
lacking. We would, however, wish to avoid both scenarios
increment increase in Hb (g/dL).
and therefore would hope for good correlation between
prediction and outcome. Our results
show a correlation coefficient of 0.64
TABLE 1. Demographic data of individuals receiving
with an R2 value of 0.35. As far as is posanalyzed transfusions
sible, we have attempted to reduce all
Variable
Number or range
confounding factors by eliminating all
Included transfusions
379
data with drain losses and colloid infuTransfusions to males
216 (57.9%)
Number of individuals
223 (mean, 1.7 per patient)
sions from the analysis. Within the
Age
1 day to 17 years 7 months (median, 6 months 15 days)
PICU, boluses of crystalloid fluids are
Weight
2.1-77.3 kg (median, 5.4 kg)
not given, while the effects of urine
Volume per kg given
2.3-47.3 mL/kg (median, 16.1 mL/kg)
output and normal maintenance fluid
214 TRANSFUSION

Volume 47, February 2007

TRANSFUSION IN CHILDREN

Fig. 3. Transfusion factor and age.

Fig. 4. Hb level before and 1 and 7 hours after transfusion.

Box = interquartile range; bars = absolute range;
line = median.

input or feeds were not analyzed as these were felt to be

normal physiologic factors.
Our derived transfusion factor of 5 has confidence
intervals within 0.28 of the estimate, which implies a small
potential difference in predicted and achieved Hb in individual patients. The higher the aimed increase in Hb concentration, the wider the potential for error between the
aimed and actual end Hb. Caution must be advised if the
need for large increments arises, which may be better
given in two smaller amounts with a check Hb value performed in between.
The weak association found between increasing age
and decreasing transfusion factor may be explained by the
differences in circulating blood volumes with age. It has
been suggested that neonates have a greater blood volume
per kilogram of body weight compared to older children,
but supporting data are limited and contradictory.4-7
The effect we describe is very small and would not be
clinically apparent. We support the advice of Morris and
coworkers8 to use of a consistent factor (we would suggest
5) for all patients whatever their age.
A questionnaire study of intensivists in Canada and
Europe showed a high variation in transfusion volume
being given to children, with most transfusions at between
10 and 20 mL per kg.9 The threshold for transfusion ranged
from 7 to 13 g per dL. Studies from adults indicate that a

more restrictive strategy (maintaining an Hb between 7

and 9) may be more beneficial, especially in the younger
than 55 years group.10 This has major implications for
the amount of blood transfused into each patient. A
recent Canadian study has suggested a transfusion rate in
children admitted to PICU of just 14 percent, although
there were certain exclusions.11 Our transfusion rate of
37.8 percent was more in keeping with that reported in a
small US study.12
We studied exclusively PICU patients and this is a
population that may have different attributes to general
patients on the wards. Most of our cardiac patients are
postoperative, and there may be small amounts of occult
blood loss. We have excluded any patients who had
overt blood loss, however, or who had enough to cause
hemodynamic compromise. The PICU population also
allows close and frequent monitoring, leading to welldocumented, robust data sets. Although it would be ideal
to repeat the study in all categories of patients, this is
unlikely to be ethical because it would need many episodes of venipuncture in a large number of children. There
is scope for further study especially in premature babies, a
population that has not been covered by our study. A
recent study assessed RBC survival in premature babies
with a different model of RBC recovery and found that the
measured values were modestly lower than those predicted, older transfused blood being associated with
poorer recovery.13
Although our study is retrospective, we have included
all charts from all patient days over 2 years, with the aim of
eradicating selection bias. We also have confidence in the
accuracy of the nursing documentation. Furthermore, our
findings are corroborated by a recently published smallscale study, which also suggested that existing formulae
underestimate the volume of RBCs required to achieve a
target Hb value and had trialled a derived factor of 4.8 in
50 critically ill children, with greater success.8
Previous anecdotal teaching that analysis of Hb just
after a transfusion would not give a reliable reading is
shown to be untrue. Hb level 7 hours after was only 0.2 g
per dL lower. This simplifies the management of patients
who need posttransfusion Hb checks and is in keeping
with the results of a recent small scale study in neonates,
which has also suggested that there is no need to wait
more than 15 minutes after a transfusion before checking
for its effect on the Hct.14 This is important as it may allow
for further administration of blood from the same unit,
before its time expiry, therefore limiting potential additional donor exposure to the child.
The blood transfused in this study was RBCs resuspended in additive solution containing saline, adenine,
glucose, and mannitol (SAG-M). A mean Hct level of close
to 0.60 has been adopted in the UK and by many transfusion centers internationally, but this may differ in various
parts of the world. Nonresuspended RBCs, which were
Volume 47, February 2007

TRANSFUSION 215

DAVIES ET AL.

REFERENCES
1. British Committee for Standards in Haematology. Transfusion guidelines for neonates and older children. Br J
Haematol 2004;124:433-53.
2. McIntosh N, Helms P, Smyth R, editors. Forfar and Arneils
textbook of pediatrics. 6th ed. London: Churchill Livingstone; 2003.
3. Nahum E, Ben-Ari J, Schonfeld T. Blood transfusion policy
among European pediatric intensive care physicians.
Fig. 5. Transfusion factor by local Hct level.

J Intensive Care Med 2004;19:38-43.

4. Samuels LD. Measuring the circulating blood: how much
does a child have? Clin Pediatr (Phila) 1972;11:479-81.

commonly transfused in the past, have an Hct level approaching 0.8. It is possible that the current transfusion
factor of 4 was calculated from blood with an Hct of 0.80
several decades ago. To correct for this difference in Hct
(and therefore in the amount of RBCs being transfused),
any local transfusion factor should be in the format transfusion constant/Hct, a lower volume of blood to be transfused if it has a higher Hct.
Extrapolation of our data suggests a possible transfusion constant which would be applicable to all Hct levels
as follows:
Transfusion factor = 3/Hct of transfused blood.
This can be demonstrated by Fig. 5.
With the UK transfused blood Hct level of 0.60 this
calculates to a transfusion factor of 5 (this would be predicted to lead to an increment of 2 g/dL when 10 mL/kg
RBCs of this Hct are transfused). Whole blood (Hct, 0.40)
would have a factor of 7.5, and blood with an Hct level of
0.80 would have a factor of 3.75. This is at present an
extrapolation and we encourage further study in units that
use differing transfusion blood Hct levels.
We present evidence-based transfusion volume calculations for children receiving RBCs with an Hct level of
0.6. Based on our results we suggest the following formula
for calculating volume of transfusion:
Volume of RBCs = Weight (kg) Increment desired
(g/dL Hb) 3/Hct.
We advise caution in large relative volume transfusions,
but confidence in transfusing up to an aimed concentration with our formula. Posttransfusion checks can be done
1 hour after transfusion instead of the current 4 to 6 hours.
In our unit we have changed practice in line with the
above formula. There is a need for a large prospective trial
in children outside of PICU to precisely delineate the predictive power of the factor of 3/Hct as a transfusion factor.

216 TRANSFUSION

Volume 47, February 2007

5. Linderkamp O, Holthausen H, Seifert J, et al. Accuracy of

blood volume estimations in critically ill children using
125I-labelled albumin and 51Cr-labelled red cells. Eur J
Pediatr 1977;125:143-51.
6. Leipala JA, Talme M, Viitala J, et al. Determination of neonatal blood volume with the haemoglobin subtype method
in appropriately grown and growth retarded preterm
infants [abstract]. Pediatr Res 1999;45:206A.
7. Aladangady N, Aitchison TC, Beckett C, et al. Is it possible
to predict the blood volume of a sick preterm infant? Arch
Dis Child Fetal Neonatal Ed 2004;89:F344-7.
8. Morris KP, Naqvi N, Davies P, Smith N, Lee PW. A new
formula for blood transfusion volume in the critically ill.
Arch Dis Child 2005;90:724-8.
9. Laverdiere C, Infante-Rivard C, Toledano B, Lacroix J.
Survey on transfusion practices of pediatric intensivists.
Pediatr Crit Care Med 2002;3:335-40.
10. Hbert P, Wells G, Blajchman M, et al. A multicenter, randomized, controlled clinical trial of transfusion requirements in critical care. N Engl J Med 1999;340:409-17.
11. Armano R, Gauvin F, Ducruet T, Lacroix J. Determinants of
red blood cell transfusions in a pediatric critical care unit:
a prospective, descriptive epidemiological study. Crit Care
Med 2005;33:2637-44.
12. Weldon BC, Conner D, Walther WJ. Red blood cell transfusion practice in critically-ill children. Crit Care Med 2003;
31(12 Suppl):A135.
13. Bechensteen AG, Chapel S, Venq-Pedersen Widness JA.
Post-transfusion recovery of stored erythrocytes in very
low birth weight infants using a hemoglobin balance
model. Transfusion 2004;44:1019-24.
14. Glatstein M, Oron T, Barak M, Mimouni FB, Dollberg S.
Posttransfusion equilibration of hematocrit in hemodynamically stable neonates. Pediatr Crit Care Med 2005;
6:707-8.