IMMUNOHEMATOLOGY
Incidence of alloantibody formation after ABO-D or extended
matched red blood cell transfusions: a randomized trial
(MATCH study)

Henk Schonewille,1,2 Aine Honohan,1,2 Leo M.G. van der Watering,1,2 Francisca Hudig,3
Peter A. te Boekhorst,4 Ankie W.M.M. Koopman-van Gemert,5 and Anneke Brand2,6
BACKGROUND: Most incidentally transfused patients
receive only ABO-Dcompatible transfusions and
antibodies are formed in up to 8%. The effect of extended
(c, C, E, K, Fya, Jka, and S antigens) matched (EM) and
ABO-Dmatched red blood cell (RBC) transfusions on
the incidence of new clinically relevant RBC antibody
formation after a first elective transfusion event in
surgical patients was studied.
STUDY DESIGN AND METHODS: A multicenter
randomized trial was performed in nontransfused
patients who were scheduled to experience a single
elective transfusion event of maximal 4 RBC units. The
primary outcome was the incidence of newly formed
warm reacting clinically relevant RBC alloantibodies
measured in three follow-up (FU) samples taken at 7 to
10 days, 4 to 6 weeks, and 4 to 6 months
posttransfusion.
RESULTS: A total of 853 patients were randomized, and
of these, 333 patients were transfused with a total of
1035 RBC units. At least one FU sample was available
from 97% of transfused patients. In intention-to-treat
analysis, new antibodies were detected in 10 of 155
ABO-D and seven of 178 EM patients, respectively.
Per-protocol analysis including 190 patients showed a
nonsignificant absolute risk difference (ARD) of 5.3%
(95% confidence interval [CI], 21.4% to 12%) in
alloimmunization between study arms. In a post hoc
analysis of 138 patients who received RBCs but no
platelet (PLT) transfusions the ARD increased to
significance, 8.0% (95% CI, 0.4-16.0).
CONCLUSION: Extended matching for selected
antigens reduced the alloimmunization risk by 64% in
surgical patients. Extended matching seems successful
only if the patient did not receive accompanying
nonmatched PLT transfusions.
C
hronically transfused patients are regularly
tested for the presence of red blood cell (RBC)
antibodies, which are detected in up to 40% of
these patients, depending on blood group dis-
parity between recipient and donor, diagnosis, treatment,
transfusion number, and genetic factors.1-5 Antibody
responders are at high risk for additional antibodies after
subsequent transfusions.6-10 Because of the high immuni-
zation rate in chronically transfused hemoglobinopathy
patients, many centers have introduced prophylactic
matching for Rhesus phenotype and K antigen. This policy
ABBREVIATIONS: ARD(s) 5 absolute risk difference(s);
ATT 5 according to transfusion; EM 5 extended matched;
FU 5 follow-up; ITT 5 intention to treat; PP 5 per
protocol.
From the 1Centre for Clinical Transfusion Research, Sanquin
Research; and the 2Jon J. Van Rood Centre for Clinical
Transfusion Research, Sanquin-Leiden University Medical
Centre, Leiden, The Netherlands; 3LabWest, Haga Teaching
Hospital, The Hague, The Netherlands; the 4Department of
Haematology, Erasmus Medical Centre, Rotterdam, The
Netherlands; the 5Department of Anesthesiology, Albert
Schweitzer Hospital, Dordrecht, The Netherlands; and the
6
Department of Immuno-Hematology and Blood Transfusion,
Leiden University Medical Centre, Leiden, The Netherlands.
Address reprint requests to: Henk Schonewille, Centre for
Clinical Transfusion Research, Sanquin Research Plesmanlaan
1a, 2333 BZ Leiden, The Netherlands; e-mail:
h.schonewille@sanquin.nl.
The study was supported by a grant from Sanquin Blood
Supply (PPOC06-017).
Received for publication June 26, 2015; revision received
August 13, 2015; and accepted August 13, 2015.
doi:10.1111/trf.13347
C 2015 AABB
V
TRANSFUSION 2016;56;311320
Volume 56, February 2016 TRANSFUSION 311
SCHONEWILLE ET AL.
has resulted in a significant, albeit variable, alloimmuniza-
tion reduction in sickle cell patients.11-15
Incidentally transfused patients receive only ABO-D
compatible RBCs; exceptions are women during (pre-)
childbearing years who in many high-income countries
receive c, E, and/or K antigenmatched transfusions.16,17
The antibody frequency in incidentally transfused patients
is not well established with reports of less than 1% up to
8%. After transfusion, many patients remain untested due
to absence of a subsequent transfusion, or a subsequent
transfusion is too soon or too late and antibodies are
either not yet or no longer detectable.18 In four studies,
performing posttransfusion longitudinal antibody follow-
up (FU), alloimmunization occurred in 4.4% to 10.5% of
patients.7,19-21
When antibodies are detected, compatible RBCs are
selected. Extensive and time-consuming tests may be nec-
essary which can delay treatment. Furthermore, for
patients with multiple antibodies, compatible blood may
not be readily available. High-throughput molecular tests
will inevitably lead to the possibility of extensive donor
genotyping. Before its implementation, evaluation of clini-
cal benefits, logistics, and costs are warranted. In particu-
lar, the question will arise whether all transfusion
recipientseven for a single transfusion eventshould
receive extended matched (EM) RBC or only antibody
responders.
In Caucasians, approximately 80% of clinically rele-
vant antibodies are directed against C, c, E, K, Fya, Jka, and
S antigens.6,22,23 We hypothesize that extended matching
for these antigens will reduce alloimmunization by 80%.
The aim of the study was to compare the effect of EM and
ABO-Dmatched RBC transfusions on the incidence of
new clinically relevant RBC antibody formation after a
first elective transfusion event in surgical patients.
MATERIALS AND METHODS
A multicenter, parallel group, randomized trial was con-
ducted in two university and two reference hospitals
between August 2008 and November 2011. Antibody FU
of the last transfused patients ended in May 2012. The
study was originally designed as a four-arm study ran-
domizing ABO-D versus EM in two patient cohorts: 1) pre-
viously nontransfused patients and 2) patients with RBC
antibodies after a previous immunizing event. However,
to be able to distinguish the immunizing transfusion epi-
sode in the second cohort, transfusion episode intervals of
at least 6 months were required, resulting in very slow
accrual of eligible patients. Accrual in this arm was dis-
continued after 1 year and not used in any analyses.
312 TRANSFUSION Volume 56, February 2016
Participants
Patients 18 years or older who were scheduled to experi-
ence a first single elective transfusion, defined as not
more than 4 RBC units within 72 hours were eligible. All
transfusion requests submitted during office hours (until
3 p.m. for logistic reasons) were assessed for eligibility by
local research nurses. Exclusion criteria included 1)
because of the lower immunization risk, an a priori indi-
cation to receive EM RBC transfusions (i.e., females
younger than 45 years, who routinely receive K-matched
RBC transfusions, hemoglobinopathy, autoimmune
hemolytic anemia), 2) patients expected to receive more
than 4 RBC units during the study transfusion episode or
3) expected to require additional transfusions within 6
months, and 4) incapacitated patients or patients who
were unable to comprehend the study information bro-
chure. Patients who received RBCs with platelet (PLT)
and/or plasma transfusion were not excluded.
Randomization
Local research nurses obtained informed consent. Per
center, patients were randomly assigned to one of two
parallel study arms, to receive either ABO-Dmatched or
EM (including c, C, E, K, Fya, Jka, and S antigens) RBC
units. Randomization was carried out centrally by the pro-
ject leader, using a computer-generated allocation table in
permuted blocks (fixed block size of 6, allocation ratio 1:1;
http://www.tufts.edu/~gdallal/random_block_size.htm).
The local hospital transfusion laboratories were informed
of the randomization result by telephone and e-mail.
Interventions
All randomized patients were typed for ABO, D, C, c, E, K,
Fya, Fyb, Jka, Jkb, M, and S blood group antigens and
screened for irregular RBC antibodies by the local hospital
transfusion laboratory with a three-cell panel using the
indirect antiglobulin test (IAT), in a low-ionic-strength
saline (LISS)-enhanced gel centrifugation technique
(DiaMed ID, Micro Typing System, DiaMed, Cressier,
Switzerland).
Either EM units were obtained from hospital stock or
a maximum of 4 units were issued through the national
(Sanquin) Blood Bank on the day preceding elective sur-
gery. Donor RBC antigen profiles are registered in the San-
quin National Donor Database and (partly) displayed on
the ISBT 128 product labels. ABO-Dmatched RBC prod-
ucts were issued from the hospitals blood bank stock.
Compatibility tests (type and screen policy) were per-
formed according to local hospital standard operating
procedures. Blood bag labeling was unchanged for study
purposes. The attending medical staff members were
unaware of the patients study participation. When RBCs
were ordered during surgery, the requested number of

units was transferred to the operating room, and any sur-
plus EM units were held for up to 72 hours postoperative.
All patients received prestorage leukoreduced RBC in
additive solution (SAG-M) with a maximum shelf life of 35
days. PLT transfusions consisted of a standard 300-mL
PLT pool, derived from five buffy coats, suspended either
in PLT AS (PAS-II) (75% of units) or in plasma (25% of
units).
Transfusion of the first unit was defined as Time Day 0
(T0); a transfusion episode was defined as units transfused
up to 72 hour after T0. The hospitals transfusion services
recorded all transfused units (RBC, PLTs, and plasma).
The trial was conducted in accordance with the prin-
ciples laid down by the Declaration of Helsinki (Version
2004). The study protocol was approved by the Medical
Ethical Committee of Southwest Holland (MEC07-099)
and of the Leiden University Medical Center (P07.262)
and the boards of directors of each participating center.
All study participants provided written informed consent.
The study was registered at the Dutch Trial Register
(www.trialregister.nl), identifier NTR1164. CONSORT 2010
guidelines were applied for reporting our parallel group
randomized trial.24
FU
Local research nurses collected three posttransfusion FU
blood samples at fixed time periods, that is, at 7 to 10 days
(FU-1), 4 to 6 weeks (FU-2), and 4 to 6 months (FU-3) to
screen for RBC antibodies. These intervals were chosen to
detect boosted, early, and later formed antibodies.18,19
During the FU visits the nurses asked the patients whether
they were transfused at nonparticipating hospitals.
Outcome
The primary outcome was the incidence of newly formed
warm reacting clinically relevant RBC alloantibodies meas-
ured in three FU samples taken at 7 to 10 days, 4 to 6 weeks,
and 4 to 6 months after a single RBC transfusion episode.
Measurements
FU samples were frozen at 2308C and antibody testing was
performed centrally at the Sanquin Research Leiden labo-
ratory by technicians blinded for randomization. All tests
used three-cell panels in three techniques with patient
sera: 1) standard test cells at room temperature for 30
minutes in NaCl gel tubes, 2) standard test cells incubated
for 15 minutes at 378C in LISS-enhanced IAT gel tubes, and
3) papain-treated test cells incubated for 15 minutes at
378C in NaCl gel tubes (DiaMed ID, Micro Typing System,
DiaMed). All positive antibody screen samples were tested
for antibody specificity using 11-cell panels, using the
same technique in which the antibody screen was positive.
Patient demographics, data with regard to all transfu-
sions (transfusion dates, products, product identification
RBC ALLOIMMUNIZATION AFTER EXTENDED MATCHING
numbers, antigen profiles), FU dates, and antibody results
were recorded on study clinical report forms. All written
information was transferred to a local secure database
(Microsoft Access 2003, Microsoft Corp., Redmond, WA).
A built-in quality system checked for irregularities, incon-
sistencies and coding errors, and clarification was
requested when necessary.
Sample size
Based on the relative frequency of C, c, E, K, Fya, Jka, and
S antibodies,6,22,23 a maximum 80% immunization risk
reduction could be obtained with matching for cognate
antigens. To detect a significant absolute reduction from
6% (based on average immunization incidence in the pro-
spective studies)7,19-21 to 1.2% (absolute risk difference
[ARD], 4.8%), two groups of 356 evaluable transfused
patients were required (power 90%, b 5 0.1). Considering
an anticipated 25% patient dropout, 445 transfused
patients were necessary in both randomization arms.
Statistical analysis
Most continuous variables were nonnormally distributed
and described as median and range, and categorical varia-
bles as number and percent. The alloimmunization risk
between the randomized groups was estimated in: 1)
intention to treat (ITT), including all randomized patients;
2) according to transfusion (ATT), transfused patients
with exclusively EM or ABO-Dmatched transfusions and
at least one antibody-FU; and 3) per-protocol (PP), includ-
ing only patients who fulfilled all study protocol criteria
analysis (i.e., 4 units in a single transfusion episode and
three FU samples). In a post hoc stratified ATT and PP
analysis the alloimmunization risk between the random-
ized groups was estimated for the subgroups of patients
who received RBCs without PLTs and for those who
received both RBC and PLTs.
Chi-square two-by-two contingency tables were used
to calculate alloimmunization risks. Results are expressed
as ARD with 95% confidence intervals (CIs).
Statistical analysis was performed with computer
software (SPSS 20.0 for Windows, SPSS, Chicago, IL). Sig-
nificance was tested using Fishers exact tests. A two-sided
p value of less than 0.05 was considered to be significant.
RESULTS
Patient enrollment
Patient enrollment commenced in August 2008 and ended
in November 2011, while FU ended in May 2012. The
study was initially designed to include all patients from
the general transfusion population. However, due to logis-
tic constraints, predominantly elective (cardiac) surgical
patients with, according to the surgical blood ordering
schedule, a high chance of transfusion support during
Volume 56, February 2016 TRANSFUSION 313
SCHONEWILLE ET AL.

Fig. 1. Flow chart showing the study course and analyses of randomized patients. ! 5 patients excluded from analysis or crossover;
pts 5 patients; EM randomized patients who exclusively received ABO-Dmatched blood (n 5 5) were crossed over to the ABO-D
arm for the ATT analysis; n-EM 5 not-EM: recipients receiving a mixture of EM and ABO-Dmatched units; T-episode 5 transfusion
episode of maximal 72 hours; n pts with/without plt-trnsf. 5 patients stratified for receipt of PLT transfusions.

surgery were eligible. In addition, slow patient accrual and
financial limitations forced early termination of patient
inclusions after 39 study months. A total of 853 patients
(female to male [F/M] ratio, 0.5) were randomized; 421
(49%) were allocated to receive ABO-D and 432 (51%) to
receive EM units (Fig. 1). Of the randomized patients, 333
(39%) were transfused. The 520 nontransfused patients
had no FU (Fig. 1) and were considered negative for RBC
antibodies in ITT analysis. Demographics and baseline
characteristics of transfused patients were balanced
between the study arms (Table 1).
ITT analysis
The 421 ABO-Drandomized patients received a total of
495 RBC units (median, 0 units; range, 0-29 units), compa-
rable to 540 RBC units (median, 0; range, 0-17) received
by the 432 EM randomized patients. In the ABO-D arm,
12 alloantibodies were detected in 10 patients (2.4%) and
10 alloantibodies in seven patients in the EM arm (1.6%;
Table 2). There was a nonsignificant ARD in alloimmuni-
zation rate of 0.8% between study arms (Fig. 2).
Transfusion and FU course
A total of 257 of 333 (77%) transfused patients complied
with the surgical blood-ordering schedule; 76 (23%)
patients either received more than 4 units within 72 hours
(n 5 34, 10%) or experienced at least one additional trans-
fusion episode (range, 1-4) within 6 months (n 5 33, 10%)
after a median of 10 days (range, 1-167 days) or experi-
enced both (n 5 9, 3%).
In the EM arm, 22 patients received 68 non-EM units
(median, 2 units; range, 1-10 units). Reasons were

314 TRANSFUSION Volume 56, February 2016

 TABLE 1. Baseline characteristics and transfusion information of patients allocated to receive ABO-D and EM
RBC units*
Allocation
Variable Total ABO-D matched EM
Randomized patients, number (% of total randomized) 853 421 (49) 432 (51)
Females/males ratio 0.5 0.6
Age (years), median (range) 68 (20-92) 67 (19-96)
Transfused patients, number (% of randomized) 333 155 (37) 178 (41)
Females/males ratio 0.7 0.6
Age (years), median (range) 70 (26-92) 69 (33-92)
Surgical patients 323 151 (97) 172 (97)
Type of surgery, number (% of surgical patients)
Cardiovascular 266 125 (83) 141 (82)
Digestive 25 9 (6) 16 (9)
Orthopedic 20 12 (8) 8 (5)
Other 12 5 (3) 7 (4)
Transfusion products in transfused patients
RBCs
Total units transfused, number (% of units) 1035 495 (48) 540 (52)
Units per patient, median (range) 2 (1-29) 2 (1-17)
Patients receiving 1-4 units 277 126 (82) 151 (85)
Patients receiving 5-9 units 42 24 (15) 18 (10)
Patients receiving 10 units 14 5 (3) 9 (5)
Patients transfused during a second episode 42 18 (12) 24 (13)
PLTs
Patients transfused 115 53 (34) 62 (35)
Total units transfused, number (% of units) 193 92 (48) 101 (52)
Units per patient, median (range) 1 (1-9) 1 (1-6)
Plasma
Patients transfused 158 74 (48) 84 (47)
Total units transfused, number (% of units) 653 318 (49) 335 (51)
Units per patient, median (range) 4 (1-18) 3 (1-15)
PLT and plasma recipients 106 49 (32) 55 (31)
* Data are expressed as number (% of allocated transfused patients), except if stated otherwise.
A unit consisted of buffy coatderived pooled PLTs from five donors.

TABLE 2. Alloantibodies detected in FU samples of ABO-D (n 5 149) and EM patients (n 5 173) tested with three
techniques*
FU sample and antibody test technique
FU-1 (7-10 days) FU-2 (4-6 weeks) FU-3 (4-6 months)
Product(s) exposing
Patient study-ID Sex LISS Enzyme LISS Enzyme LISS Enzyme cognate antigens
ABO-Dmatched arm (total 12 antibodies)
30 Male E E E E PLTs (1)
47 Female C RBCs (7) 1 PLTs (7)
128 Male K K RBCs (2)
131 Female Jka Jka Jka Jka RBCs unknown
161 Female E RBCs (1)
166 Female E c1E c1E RBCs (2, 1)
168 Male E1K E1K RBCs (1, 1)
198 Male K RBCs (1)
230 Female K RBCs (1)
267 Female NA NA Cw Cw Cw Cw RBCs/PLTs unknown
All antibodies 1 1 3 4 9 9
EM arm (total 10 antibodies)
26 Male E PLTs (4)
55 Female NA NA E E PLTs (5)
219 Female E 1 Jka RBCs (2, 7)
238 Male NA NA E1K E1K PLTs (9, 1)
273 Male Lua Lua Lua Lua RBCs unknown
277 Male e e PLTs (4)
294 Male D1E D1E D D PLTs (5, 2)
All antibodies 0 0 3 4 6 8
* The numbers in parentheses in the last column represent the number of donors with the cognate antigens. Patients with antibodies included
for ATT and PP analyses are shown in Table S2A. NA 5 no FU sample available; for layout purposes negative and reactions at room tem-
perature (of which one of 885 FU samples showed nonspecific test reactions) were not depicted; enzyme 5 papain-treated test cells; RBCs
unknown, these two patients had only received RBCs and donor Jka and Lua antigens were not typed, respectively; RBCs/PLTs unknown,
this patient had received RBCs and PLTs and Cw antigen of all six donors involved was not typed.
SCHONEWILLE ET AL.
Fig. 2. Forest plot of ARD in alloimmunization rate between patients receiving ABO-Dmatched and EM RBCs; ITT, ATT, PP, and
subanalyses after stratification for PLT transfusions.
incorrect ordering of nonmatched or partly EM matched
units (protocol violations, eight patients) or the receipt of
more than 4 units within 72 hours or more than one trans-
fusion episode (14 patients). Overall, 156 EM patients
(88%) received only EM units and 94.4% of the transfu-
sions were according to protocol. Except for a slightly
higher number of RBC units transfused beyond 72 hours
in the EM patients, the transfusion course in randomiza-
tion arms was comparable (Fig. 1 and Table 1).
Adherence to FU sampling was comparable for ran-
domization arms and complete for 238 patients (71.5%),
incomplete (one or two missing samples) for 84 patients
(25.2%), and absent for 11 patients (3.3%; Table S3, avail-
able as supporting information in the online version of
this paper). For 87.0% of samples, FU was obtained within
the defined period: FU-1, 81%; FU-2, 88%; and FU-3, 93%.
Alloantibody incidence rate increased with longer FU;
in 294 FU-1 samples one (0.3%) antibody, in 290 FU-2
samples eight (2.8%) antibodies, and in 270 FU-3 samples
19 (6.8%) antibodies. After initial detection, three antibod-
ies (anti-Jka and 2 anti-E) were no longer detectable at last
FU (Table 2).
ATT analysis
The five EM patients who received only non-EM units were
crossed over to the ABO-Dmatched arm. The 11 patients
without any FU and 17 EM patients who received partly
non-EM units were excluded, resulting in 154 ABO-D and
151 EM evaluable patients (Fig. 1). Baseline transfusion
and FU characteristics were balanced between the study
arms (Tables S1A and S3A, available as supporting informa-
tion in the online version of this paper). In the ABO-D arm,
316 TRANSFUSION Volume 56, February 2016
12 RBC alloantibodies were detected in 10 patients (6.5%)
and seven alloantibodies in five patients in the EM arm
(3.3%; Table S2A, available as supporting information in
the online version of this paper). There was a nonsignifi-
cant ARD in alloimmunization of 3.1% (95% CI, 21.7 to
8.0) between study arms (Fig. 2).
PP analysis
All study protocol criteria were fulfilled by 190 transfused
patients (57%; 86 ABO-Dmatched and 104 EM; Fig. 1).
Baseline transfusion and FU characteristics were balanced
between the study arms (Tables S1B and S3B, available as
supporting information in the online version of this
paper). In the ABO-D arm seven patients (8.1%) formed
nine alloantibodies compared to four alloantibodies in
three EM patients (2.9%; Table S2A). There was a nonsigni-
ficant ARD in alloimmunization rate of 5.3% (95% CI, 21.4
to 12.0) between study arms (Fig. 2).
Alloantigen exposure by RBC and PLT
transfusions
The cumulative number of D, C, c, E, K, Fya, Jka, and S
antigens which were not expressed by 322 patients with at
least one FU was 1066 (median per patient, 3; range, 1-6).
Of these, 135 patients were not exposed to mismatched
antigens, 146 patients were exposed to 278 mismatched
antigens, and for 41 patients exposure to 51 antigens
could not be ascertained because not all units were
completely typed. Table 3 shows the association
between antigenic exposure and antibody formation.
Immunization risk after RBC transfusion varied from
2.6% to 12.9% for the various antigens. Six of the 17
 RBC ALLOIMMUNIZATION AFTER EXTENDED MATCHING

TABLE 3. Risk on alloantibody formation after exposure to study antigens by RBC transfusion in the 322 patients
not expressing a given antigen and with at least one antibody FU*
Alloantibody
formed
Patients not Antigen exposure
Antigen expressing antigen by RBC transfusion No Yes IR (%) 95% CI
D 44 (14) Yes 0 0
No 43 1 NA
C 88 (27) Yes 28 1 3.4 0.2-19.6
No 59 0
c 67 (21) Yes 37 1 2.6 0.1-15.4
No 29 0
E 238 (74) Yes 63 4 6.0 1.9-15.4
No 166 5 NA
K 297 (92) Yes 27 4 12.9 4.2-30.8
No 260 1 NA
Unknown 5 0
Fya 109 (34) Yes 36 0
No 51 0
Unknown 22 0
Jka 67 (20) Yes 26 1 3.7 0.2-20.9
No 35 0
Unknown 4 1 NA
S 156 (48) Yes 50 0
No 87 0
Unknown 19 0
* Data are reported as number (%). Unknown 5 antigen type was not known for RBC units; IR (%) 5 immunization risk (% patients with anti-
bodies as fraction of all cognate antigen exposed patients).
These antibodies were formed after Rh- and K-mismatched PLT transfusions.
IR changed to 11.1% (4/36; 95% CI, 3.6-27.0) for K and to 6.3% (2/32; 95% CI 1.1-22.2) for Jka when the patients with unknown exposure
(n 5 5 for both antigens) were included in the calculations.

alloimmunized patients had formed eight antibodies
(five anti-E, one -D, -K, and -e) without being exposed
to cognate antigens by RBC transfusions. To investigate
the involvement of PLT transfusion on RBC alloantibody
formation, the Rh and K antigens of all PLT donors
involved in the study were retrieved. A total of 80
patients were exposed to Rh and K antigens by PLT
transfusions only. For all eight antibodies, the cognate
antigens were expressed by the PLT donor only and
were considered the immunizing source (Table 2).
Post hoc stratified ATT and PP analysis
ATT analysis
Of the 201 patients who received only RBC transfusions,
7.1% ABO-Dmatched patients formed antibodies com-
pared to 1.0% EM patients resulting in a significant ARD
in alloimmunization of 6.1% in favor of extended match-
ing. The alloimmunization risk among the 104 patients
who received PLT transfusions in addition to RBCs was
5.5% for the ABO-Dmatched patients versus 8.2% for the
EM patients (p 5 0.7, Fig. 2).
PP analysis
Of the 138 patients who received not more than 4 RBC
transfusions according to protocol but no PLT transfusions
and with complete FU, 9.4% of ABO-Dmatched patients
formed antibodies compared to 1.4% of EM patients result-
ing in a significant ARD in alloimmunization of 8.0% in
favor of extended matching. The alloimmunization risk
among the 52 patients who received PLT transfusions in
addition to RBCs was 4.5% for the ABO-Dmatched
patients versus 6.7% for the EM patients (p 5 1.0, Fig. 2).
DISCUSSION
In this randomized controlled trial, comparing alloimmu-
nization after ABO-D and EM RBC transfusions, ITT, ATT,
and PP analyses showed nonsignificant ARDs in RBC
alloimmunization between the groups. Three patients
(one in the EM and two in the ABO-D group) formed allo-
antibodies against nonmatched antigens (anti-Cw, -Lua,
and -e). Surprisingly, among the patients who received
PLT transfusions in addition to RBCs, six patients formed
RBC alloantibodies after RBC antigen exposure through
PLT transfusions only. Moreover, a significant ARD in favor
of extended RBC matching was shown in both ATT and PP
analysis, 6 and 8%, respectively (relative risk reduction of
85%) only in the group of patients who did not receive
PLT transfusions. These results indicate that RBC
extended matching is only successful if the patient does
not receive accompanying nonRBC antigenmatched
PLT products.
Antibodies were detected with increasing frequency
at later FU time points. At FU-1, 5%, at FU-2, 36%, and at

Volume 56, February 2016 TRANSFUSION 317

SCHONEWILLE ET AL.
FU-3, 86% of all 22 alloantibodies, detected during the
total FU period, were present. Three antibodies did not
persist through to the last FU. As the median period
between FU-2 and FU-3 was approximately 4 months, for-
mation and disappearance of new antibodies within this
period cannot be ruled out. Furthermore, as FU ended
after a median of 5.2 months antibodies may have been
formed thereafter. Given the results from two prospective
studies with a longitudinal FU up to 9 and 15 months
respectively, both showing all new antibodies were formed
within 6 months after transfusion, this is unlikely.19,20
To our knowledge, this is the first randomized post-
transfusion antibody FU study in which antigen profiles of
both recipients and donors were available, making it pos-
sible to investigate the immunization risk and kinetics per
mismatched RBC antigen and, although not intentionally
planned, to estimate the risk of alloimmunization due to
accompanying PLT transfusions. Because attending medi-
cal staff members were unaware of their patients partici-
pation, the study reflects daily transfusion practice in
surgical patients, considered eligible for transfusion.
The study also has a number of limitations. Shortly
after study initiation it appeared that the design was too
problematic for nonsurgical patients needing often urgent
and not elective transfusions, and compatible units may
not be readily available in hospital blood bank stock.
Although the study design could not exclude primary
immunization during pregnancies, in only one female
patient a transient antibody was detected during the first
FU, making previous immunizations influencing the
results unlikely.
Antibody FU was performed three times at fixed
intervals after transfusion. Antibodies may become unde-
tectable over time; therefore, we cannot exclude that
immunization rate would be higher if screening was per-
formed with shorter time intervals.25-28
Due to the complexity of patient accrual, prolonging
the inclusion time, and financial limits the study was ter-
minated prematurely, weakening the power and widening
the CIs. Finally, our study included a large population of
nonimmunocompromised cardiac surgery patients and
therefore our results may not be generalizable to other
patient populations.
In our study only patients experiencing a first transfu-
sion event were included. Although the aforementioned
four prospective studies on RBC antibody formation did
not exclude patients with a transfusion history evoking a
possible memory response, the antibody incidence and
kinetics we observed show striking similarities.7,19-21 Over-
all new clinically relevant alloimmunization incidences in
these four studies varied from 4.4% to 8.4% of which
approximately one-third had multiple specificities. We
observed, after a comparable number of RBC units (2-4)
and FU interval, an overall new clinically relevant alloim-
munization incidence of 6.5% in the ATT analysis with
318 TRANSFUSION Volume 56, February 2016
29% multiantibody responders in our ABO-Dmatched
study arm. A minority of antibodies was formed within 8
weeks and the remaining between 2 and 6 months. Fur-
thermore, a substantial proportion of these antibodies
appearing within 8 weeks have disappeared at 6 months.
This was the case in three of 10 antibodies in the UK
PRISM A study and in three of the eight antibodies in our
study.21 Combining all study results, it may be concluded
that in nonimmunocompromised patients the majority of
primary response antibodies do not reach detectable lev-
els within 2 to 3 months and that antibody formation
beyond 6 months after transfusion is less likely. This out-
come questions published data on alloimmunization inci-
dence, which lack adequate FU intervals for antibody
detection.
Enzyme-treated test RBCs are more sensitive to
detect Rhesus and Kidd antibodies but, because these
unlikely cause a hemolytic reaction, are not routinely used
in pretransfusion antibody screening tests. We used
enzyme-treated cells primarily to investigate whether
these antibodies preceded detection using the LISS tech-
nique, as shown by Redman and colleagues.19 In only one
patient (ID 166) anti-E was detected in the enzyme
method before anti-E plus anti-c became detectable in
the LISS method. During the last FU, however, four anti-
bodies in three patients (ID 47, 161, and 219) were
detected for the first time only in the enzyme test, and it
cannot be ruled out that these antibodies would become
detectable later with the LISS method, and thus we
defined these antibodies as clinically relevant.
Despite 120 patients being exposed to one or more
Fya, Jka, and/or S antigens, only two patients formed anti-
Jka. This may either reflect the relative low immunogenic-
ity of these antigens or the median 5-month FU period
was perhaps too short after all. In other posttransfusion
antibody FU studies these specificities were encountered
with frequencies varying from 0, 8, 15, to 35% (combined
frequency, 14%).7,19-21 In the study with the highest pro-
portion of these specificities almost 30% of the patients
had a transfusion history suggesting that an antigen
booster is necessary to reach detectable antibody levels to
these antigens.7 Regarding Rh and K antibodies, we
observed that in the majority of responding patients, 1 to
2 incompatible units (RBCs or PLTs) were sufficient to
induce antibodies, confirming the relative high immuno-
genicity of these antigens (D. Evers, R.A. Middelburg,
S. Zalpuri, et al., submitted for publication).29
Residual nonD-matched RBC in PLT products have
been reported to be the offender of anti-D in mainly
immunosuppressed hematooncology patients, with
reports of 0% to 14%,30-32 while non-D antibodies have
been scarcely reported.33-35 In this study seven non-D
antibodies (five anti-E, one anti-e and anti-K) were
formed by six EM patients after cognate antigen exposure
through (nonmatched) PLT transfusions. As only one of

these patients was female and none of these antibodies
were detected in the available first FU samples, a post-
pregnancy booster response or passively (plasma)
acquired antibodies was unlikely. The standard 300-mL
PLT pool, derived from five buffy coats, contains less than
5 3 109 RBCs.17 This implies that, in responding recipi-
ents, primary non-D, Rh, and K antibody formation can
be induced by less than 0.5 mL of RBCs.
High-throughput systems to genotype donors and
patients will enable RBC matching for selected patients.
In the current and UK PRISM A study less than 50% of
randomized surgical patients were actually transfused
implying unnecessary hospital work and costs for these
patients.21 Furthermore, approximately 25% of patients
required more than the expected number of RBC units
during surgery. Most striking, to obtain maximal effect of
RBC extended matching it is also necessary to match
accompanying PLT transfusions for at least Rh and K anti-
gens. Considering that five donors are involved in buffy
coatderived PLTs, logistic and stocking problems would
be imminent, but might be less with the use of single-
donor apheresis PLTs. The latter product is also associated
with less RBC contamination and may result in a lower
RBC immunization risk.31
In conclusion, preemptive extended matching for
selected antigens reduced the RBC alloimmunization risk
by 64% in surgical first-time RBC recipients. Extended
RBC matching seems to be successful only if the patient
does not receive accompanying nonRBC antigen
matched PLT transfusions. This unexpected finding
should be further assessed and given consideration in
costbenefit analysis for implementation of a preemptive
extended matching strategy in the various patient
populations.
ACKNOWLEDGMENTS
We are grateful to all laboratory technicians and nurses involved
in the study at the participating hospitals. Our thanks go to Trudy
van Boxtel-Groeneveld, Sandra Mollee-Huurdeman, Pauline
Lindhout, Marjan van der Elst, Corine van Tricht, Florence Sard-
joe, Annette van Zijl (), Marijke Verbaan, Karin Prinsen-Zander,
Michaela van Bohemen, Cees Scherpenisse, Bea Koetsier, and
Yvonne Dubbeldam for their help in the recruitment of patients,
laboratory work, and collection of FU samples.
CONFLICT OF INTEREST
The authors have disclosed no conflicts of interest.
REFERENCES
1. Bauer MP, Wiersum-Osselton J, Schipperus M, et al. Clinical
predictors of alloimmunization after red blood cell transfu-
sion. Transfusion 2007;47:2066-71.
RBC ALLOIMMUNIZATION AFTER EXTENDED MATCHING
2. Zalpuri S, Zwaginga JJ, le Cessie S, et al. Red-blood-cell
alloimmunization and number of red-blood-cell transfu-
sions. Vox Sang 2012;102:144-9.
3. Sanz C, Nomdedeu M, Belkaid M, et al. Red blood cell
alloimmunization in transfused patients with myelodysplas-
tic syndrome or chronic myelomonocytic leukemia. Transfu-
sion 2013;53:710-5.
4. Zalpuri S, Evers D, Zwaginga JJ, et al. Immunosuppressants
and alloimmunization against red blood cell transfusions.
Transfusion 2014;54:1981-7.
5. Schonewille H, Doxiadis II, Levering WH, et al. HLA-DRB1 asso-
ciations in individuals with single and multiple clinically rele-
vant red blood cell antibodies. Transfusion 2014;54:1971-80.
6. Tormey CA, Fisk J, Stack G. Red blood cell alloantibody fre-
quency, specificity, and properties in a population of male
military veterans. Transfusion 2008;48:2069-76.
7. van de Watering L, Hermans J, Witvliet M, et al. HLA and
RBC immunization after filtered and buffy coat-depleted
blood transfusion in cardiac surgery: a randomised con-
trolled trial. Transfusion 2003;43:765-71.
8. Schonewille H, van de Watering LM, Brand A. Additional red
blood cell alloantibodies after blood transfusions in a non-
hematologic alloimmunized patient cohort: is it time to take
precautionary measures? Transfusion 2006;46:630-5.
9. Schonewille H, Klumper FJ, van de Watering LM, et al. High
additional maternal red cell alloimmunization after Rhesus-
and K-matched intrauterine intravascular transfusions for
hemolytic disease of the fetus. Am J Obstet Gynecol 2007;
196:143.e1-6.
10. Schonewille H, de Vries RR, Brand A. Alloimmune
response after additional red blood cell antigen challenge
in immunized hematooncology patients. Transfusion
2009;49:453-7.
11. Rosse WF, Gallagher D, Kinney TR, et al. Transfusion and
alloimmunization in sickle cell disease. The Cooperative
Study of Sickle Cell Disease. Blood 1990;76:143127.
12. Vichinsky EP, Luban NL, Wright E, et al. Prospective RBC
phenotype matching in a stroke-prevention trial in sickle
cell anemia: a multicenter transfusion trial. Transfusion
2001;41:1086-92.
13. LaSalle-Williams M, Nuss R, Le T, et al. Extended red blood
cell antigen matching for transfusions in sickle cell disease:
a review of a 14-year experience from a single center (CME).
Transfusion 2011;51:1732-9.
14. Miller ST, Kim HY, Weiner DL, et al. Red blood cell alloim-
munization in sickle cell disease: prevalence in 2010. Trans-
fusion 2013;53:704-9.
15. Chou ST, Jackson T, Vege S, et al. High prevalence of red
blood cell alloimmunization in sickle cell disease despite
transfusion from Rh-matched minority donors. Blood 2013;
122:1062-71.
16. Goldman M, Lane D, Webert K, et al. The prevalence of anti-K
in Canadian prenatal patients. Transfusion 2015;55:1486-91.
17. English translation of the Dutch blood transfusion guideline
2011 [Internet]. Amsterdam: ISBT Central Office; 2012 [cited
Volume 56, February 2016 TRANSFUSION 319
SCHONEWILLE ET AL.
July 2014]. Available from: http://www.diliguide.nl/docu-
ment/7584
18. Schonewille H, van de Watering LM, Loomans DS, et al. Red
blood cell alloantibodies after transfusion: factors influenc-
ing incidence and specificity. Transfusion 2006;46:250-6.
19. Redman M, Regan F, Contreras M. A prospective study of the
incidence of red cell allo-immunisation following transfu-
sion. Vox Sang 1996;71:216-20.
20. Alves VM, Martins PR, Soares S, et al. Alloimmunization
screening after transfusion of red blood cells in a prospective
study. Rev Bras Hematol Hemoter 2012;34:206-11.
21. Elebute MO, Choo L, Mora A, et al. Transfusion of prion-
filtered red cells does not increase the rate of alloimmuniza-
tion or transfusion reactions in patients: results of the UK
trial of prion-filtered versus standard red cells in surgical
patients (PRISM A). Br J Haematol 2013;160:701-8.
22. Schonewille H, Brand A. Alloimmunization to red blood cell
antigens after universal leucodepletion. A regional multi-
centre retrospective study. Br J Haematol 2005;129:151-6.
23. Winters JL, Richa EM, Bryant SC, et al. Polyethylene glycol anti-
globulin tube versus gel microcolumn: influence on the inci-
dence of delayed hemolytic transfusion reactions and delayed
serologic transfusion reactions. Transfusion 2010;50:1444-52.
24. Schultz KF, Altman DG, Moher D, et al. CONSORT 2010
statement: updated guidelines for reporting parallel group
randomised trials. J Clin Epidemiol 2010;63:834-40.
25. Ramsey G, Smietana SJ. Long-term follow-up testing of red
cell alloantibodies. Transfusion 1994;34:122-4.
26. Schonewille H, Haak HL, van Zijl AM. RBC antibody persist-
ence. Transfusion 2000;40:1127-31.
27. Reverberi R. The persistence of red cell alloantibodies. Blood
Transfus 2008;6:225-34.
28. Tormey CA, Stack G. The persistence and evanescence of
blood group alloantibodies in men. Transfusion 2009;49:
505-12.
29. Hogervorst E, Middelburg R, Brand A, et al. Anti-K formation
is not associated with the storage time of transfused red
blood cells. Transfusion 2015;55(6 Pt 2):1472-7.
30. Atoyebi W, Mundy N, Croxton T, et al. Is it necessary to
administer anti-D to prevent RhD immunization after the
transfusion of RhD-positive platelet concentrates? Br J Hae-
matol 2000;111:980-3.
 G, Pereira A, et al. Platelet transfusions from
31. Cid J, Carbasse
D1 donors to D- patients: a 10-year follow-up study of 1014
patients. Transfusion 2011;51:1163-9.
320 TRANSFUSION Volume 56, February 2016
32. OBrien K, Haspel RL, Uhl L. Anti-D alloimmunisation
after D-incompatible platelet transfusions: a 14-year sin-
gle-institution retrospective review. Transfusion 2014;54:
650-4.
33. Kitazawa J, Nollet K, Morioka H, et al. Non-D Rh antibodies
appearing after apheresis platelet transfusion: stimulation by
red cells or microparticles? Vox Sang 2011;100:395-400.
34. Moncharmont P, Barday G, Meyer F. Red blood cell alloim-
munisation after platelet transfusion: a 5-year study. Blood
Transfus 2014;12Suppl1:s147-8.
35. Moncharmont P, Barday G, Meyer F. Red blood cell alloim-
munisation in 18 to 50-year old transfused women: a 3-year
study. Blood Transfus 2015;13:345-6.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the
online version of this article at the publishers website:
Table S1A. Baseline characteristics and transfusion
information of patients allocated to receive ABO-D and
extended matched RBC units for according-to-
transfusion analyses.
Table S1A-1. Baseline characteristics and transfusion
information of patients allocated to receive ABO-D and
extended matched RBC units for according-to-
transfusion analyses, stratified for patients receiving
RBC and platelets and patients receiving RBC only.
Table S1B. Baseline characteristics and transfusion
information of patients allocated to receive ABO-D and
extended matched RBC units for Per-protocol analyses.
Table S1B-1. Baseline characteristics and transfusion
information of patients allocated to receive ABO-D and
extended matched RBC units for per-protocol analyses,
stratified for patients receiving RBC and platelets and
patients receiving RBC only.
Table S2A. Alloantibodies detected in follow-up samples
of ABO-D and extended matched patients tested with
three techniques, and included for according-to-
transfusion (ATT) and per-protocol analyses.
Table S3. Follow-up sampling after the first RBC unit
(T1) in ABO-D and extended matched patients
Table S3A. Follow-up sampling after the first RBC unit
(T1) in ABO-D and extended matched patients,
included for the according-to-transfusion and per-
protocol analyses.