ARTICLE Myelodysplastic Syndrome

Ferrata Storti Foundation Impact of red blood cell transfusion dose

density on progression-free survival in patients
with lower-risk myelodysplastic syndromes
Louise de Swart,1 Simon Crouch,2 Marlijn Hoeks,1,3 Alex Smith,2 Saskia
Langemeijer,1 Pierre Fenaux,4 Argiris Symeonidis,5 Jaroslav Čermák,6 Eva
Hellström-Lindberg,7 Reinhard Stauder,8 Guillermo Sanz,9 Moshe Mittelman,10
Mette Skov Holm,11 Luca Malcovati,12 Krzysztof Mądry,13 Ulrich Germing,14
Aurelia Tatic,15 Aleksandar Savic,16 Antonio Medina Almeida,17 Njetočka
Haematologica 2020 Gredelj-Šimec,18 Agnes Guerci-Bresler,19 Odile Beyne-Rauzy,20 Dominic
Volume 105(3):632-639 Culligan,21 Ioannis Kotsianidis,22 Raphael Itzykson,4 Corine van Marrewijk,1
Nicole Blijlevens,1 David Bowen23 and Theo de Witte,24 on behalf of the EUMDS
Registry Participants
1
Department of Hematology, Radboud University Medical Center, Nijmegen, the
Netherlands; 2Epidemiology and Cancer Statistics Group, Department of Health Sciences,
University of York, York, UK; 3Center for Clinical Transfusion Research, Sanquin Research,
Leiden, the Netherlands; 4Service d'Hématologie, Hôpital Saint-Louis, Assistance
Publique des Hôpitaux de Paris and Université Paris 7, Paris, France; 5Department of
Medicine, Division of Hematology, University of Patras Medical School, Patras, Greece;
6
Department of Clinical Hematology, Institute of Hematology & Blood Transfusion, Praha,
Czech Republic; 7Department of Medicine, Division of Hematology, Karolinska Institutet,
Stockholm, Sweden; 8Department of Internal Medicine V (Hematology and Oncology),
Innsbruck Medical University, Innsbruck, Austria; 9Department of Hematology, Hospital
Universitario y Politécnico La Fe, Valencia, Spain; 10Department of Medicine A, Tel Aviv
Sourasky (Ichilov) Medical Center and Sackler Medical Faculty, Tel Aviv University, Tel Aviv,
Israel; 11Department of Hematology, Aarhus University Hospital, Aarhus, Denmark;
12
Department of Hematology Oncology, Fondazione IRCCS Policlinico San Matteo,
University of Pavia, Pavia, Italy; 13Department of Hematology, Oncology and Internal
Medicine, Warsaw Medical University, Warsaw, Poland; 14Department of Hematology,
Oncology and Clinical Immunology, Universitätsklinik Düsseldorf, Düsseldorf, Germany;
15
Center of Hematology and Bone Marrow Transplantation, Fundeni Clinical Institute,
Bucharest, Romania; 16Clinic of Hematology - Clinical Center of Vojvodina, Faculty of
Correspondence: Medicine, University of Novi Sad, Novi Sad, Serbia; 17Department of Clinical Hematology,
Hospital da Luz, Lisbon, Portugal; 18Department of Internal Medicine, Division of
Hematology, Merkur University Hospital, Zagreb, Croatia; 19Service d'Hématologie, Centre
THEO DE WITTE
Hospitalier Universitaire (CHU) Brabois Vandoeuvre, Nancy, France; 20Service de
[email protected]
Médecine Interne, IUCT-Oncopole, CHU Toulouse, Toulouse, France; 21Department of
Haematology, Aberdeen Royal Infirmary, Aberdeen, UK; 22Department of Hematology,
Democritus University of Thrace Medical School, University Hospital of Alexandroupolis,
Received: November 19, 2018.
Accepted: June 5, 2019. Alexandroupolis, Greece; 23St. James's Institute of Oncology, Leeds Teaching Hospitals,
Leeds, UK and 24Department of Tumor Immunology - Nijmegen Center for Molecular Life
Sciences, Radboud University Medical Center, Nijmegen, the Netherlands
Pre-published: June 6, 2019.

doi:10.3324/haematol.2018.212217
ABSTRACT
Check the online version for the most updated
information on this article, online supplements,

P
and information on authorship & disclosures: rogression-free survival (PFS) of patients with lower-risk myelodys-
www.haematologica.org/content/105/3/632 plastic syndromes (MDS) treated with red blood cell transfusions is
usually reduced, but it is unclear whether transfusion dose density is
an independent prognostic factor. The European MDS Registry collects
©2020 Ferrata Storti Foundation prospective data at 6-monthly intervals from newly diagnosed lower-risk
Material published in Haematologica is covered by copyright. myelodysplastic syndromes patients in 16 European countries and Israel.
All rights are reserved to the Ferrata Storti Foundation. Use of
published material is allowed under the following terms and
Data on the transfusion dose density - the cumulative dose received at the
conditions: end of each interval divided by the time since the beginning of the interval
https://creativecommons.org/licenses/by-nc/4.0/legalcode. in which the first transfusion was received - were analyzed using propor-
Copies of published material are allowed for personal or inter-
nal use. Sharing published material for non-commercial pur-
tional hazards regression with time-varying co-variates, with death and
poses is subject to the following conditions: progression to higher-risk MDS/acute myeloid leukemia as events. Of the
https://creativecommons.org/licenses/by-nc/4.0/legalcode, 1,267 patients included in the analyses, 317 died without progression; in
sect. 3. Reproducing and sharing published material for com-
mercial purposes is not allowed without permission in writing
162 patients the disease had progressed. PFS was significantly associated
from the publisher. with age, EQ-5D index, baseline World Health Organization classification,
bone marrow blast count, cytogenetic risk category, number of cytopenias,
and country. Transfusion dose density was inversely associated with PFS

632 haematologica | 2020; 105(3)

 Transfusion dose density in lower-risk MDS

(P<1x10-4): dose density had an increasing effect on hazard until a dose density of 3 units/16 weeks. The
transfusion dose density effect continued to increase beyond 8 units/16 weeks after correction for the impact
of treatment with erythropoiesis-stimulating agents, lenalidomide and/or iron chelators. In conclusion, the
negative effect of transfusion treatment on PFS already occurs at transfusion densities below 3 units/16
weeks. This indicates that transfusion dependency, even at relatively low dose densities, may be considered
as an indicator of inferior PFS. This trial was registered at www.clinicaltrials.gov as #NCT00600860.

Introduction Methods
Red blood cell transfusions (RBCT) are the major com- Patients with lower-risk MDS (IPSS risk: low or intermediate-1)3
ponent of the supportive care of patients with myelodys- from 16 European countries and Israel were included in the
plastic syndromes (MDS). The life expectancy of MDS EUMDS Registry, after providing signed informed consent, within
patients treated with RBCT is usually shorter than that of 100 days of their initial diagnosis of a MDS which was made
untransfused patients,1,2 but whether the impaired out- according to the World Health Organization (WHO) 2001 crite-
come is a result of intrinsic deterioration of the underlying ria.18 Patients with an IPSS intermediate-2 or high risk, or with
disease or a result of external factors related to transfusion therapy-related MDS were excluded, but MDS-specific treatment,
per se (for example the iron toxicity induced by RBCT) started before registration within 100 days after diagnosis, was not
remains an open question. Since 2007, the European MDS a reason for exclusion. Data were collected at baseline and at each
(EUMDS) Registry has prospectively collected observa- 6-monthly outpatient routine follow-up visit. Clinical information
tional data on patients with MDS classified as low and was collected on: demographics, anthropometrics, co-morbidities,
intermediate-1 risk according to the International performance status, quality of life (EQ-5D), concomitant medica-
Prognostic Scoring System (IPSS),3 collectively defined as tion, laboratory parameters, diagnostics including information on
lower-risk MDS.4 The majority of lower-risk MDS patients bone marrow morphology, histology, cytogenetics, RBCT
become transfusion dependent (51% in the EUMDS episodes, total number of transfused units and simultaneous ther-
Registry),4 usually within 6 months after diagnosis. With apeutic interventions. All subjects were followed prospectively by
an expected median survival of 2.4 to 11.8 years, these full reports every 6 months until death, progression to high risk
patients might be prone to long-term accumulation of iron MDS or leukemia, loss to follow-up or withdrawal of informed
due to RBCT.3,5-8 The toxic effects of iron overload in other consent. The Registry was approved by each institution’s ethics
iron-loading diseases, such as hereditary hemochromato- committee in accordance with national legislation.
sis9 and the thalassemia syndromes,10 are well known, but Transfusion data available in the EUMDS Registry consists of
the consequences in MDS patients require further clarifica- the number of units received between each reported visit, usually
tion. MDS patients are generally older than patients with at 6-month intervals. In order to assess the association between
other iron-loading disorders.11 Their exposure to RBCT transfusions received and PFS, proportional hazards regression
may not be long enough to develop classical tissue damage with time-varying covariates was employed, adjusting the effect
due to iron overload, but they may suffer from oxidative of transfusions by appropriate baseline and time-varying variables.
stress caused by toxic iron species, including non-transfer- For the purposes of the time-to-event analyses, time was meas-
rin bound iron (NTBI) and labile plasma iron (LPI), which ured from the date of diagnosis with MDS to the date of disease
have been suggested to serve as early indicators of iron progression or date of death. Progression is defined as an increase
toxicity in iron-loading anemias, such as thalassemia syn- to either refractory anemia with excess blasts-2 or to acute
dromes.8,12 Biomarkers of oxidative stress have been found leukemia. Patients without disease progression and still alive at
to be increased in patients with MDS and iron overload.4,13- the time of the analyses were censored at the date of their last
16
Data from a recently completed study of the EUMDS visit.
Registry17 showed that elevated LPI levels - in contrast to In order to avoid problems with simultaneity of cause and effect
elevated NTBI levels and transferrin saturation - are asso- assumed by the proportional hazards approach to survival analy-
ciated with decreased survival. The risk of dying prema- sis a “dose density” variable was defined, in the following way, for
turely in patients with detectable LPI levels occurred too blood transfusions received. The cumulative total of units of blood
early in this study to be explained by classical iron over- received at the end of each inter-visit time interval was calculated.
load with organ toxicity (lungs, liver and heart) after long This was then divided by the time since the beginning of the time
term transfusions, and this may suggest a direct effect interval in which the first post-diagnosis transfusion was received,
associated with elevated LPI levels. giving a dose-density measurement. This dose-density was then
The aim of this analysis was to assess the effect of assigned to each time interval. The value of this variable at each
RBCT dose density on progression-free survival (PFS) of point in time represents the average rate at which the patient has
patients with lower-risk MDS. The hypothesis is that been receiving units of blood since they started transfusions.
transfusional iron may be toxic and associated with oxida- Adjusted baseline variables included age at diagnosis, number
tive stress, which may lead to bone marrow failure, genet- of cytopenias and number of units of blood received before diag-
ic damage, increased risk for progression or premature nosis. Adjusted time-varying variables (with the intention of
death. Two countervailing forces may play a role in this adjusting for the condition of the patient over time) were bone
analysis: (i) patients with symptomatic anemia are more marrow blast count, EQ-5D index, revised IPSS cytogenetic cate-
likely to receive more frequent RBCT; and (ii) higher gory, and platelet and neutrophil counts. Additional analyses were
RBCT doses may lead to faster deterioration of lower-risk adjusted for the effect of treatment with erythropoiesis-stimulat-
MDS or to a higher risk of complications by co-morbidi- ing agents (ESA), iron chelation therapy and lenalidomide, taking
ties. these treatments to be confounding factors. Finally, a sensitivity

haematologica | 2020; 105(3) 633

 L. de Swart et al.

analysis was performed in the survival regressions to take into 14, 2017 of whom 1,504 patients had data recorded from
account that the population was not homogeneous but distributed three or more visits (visit 3 = landmark at the 1-year fol-
over different centers in several countries, using a random effects low-up). Two patients with refractory anemia with excess
frailty term. The random effect, called “frailty”, is the term that blasts-2 were excluded, resulting in the inclusion of 1,502
describes the common risk or the individual heterogeneity, acting patients. An additional 235 patients were excluded, as one
as a factor on the hazard function. Missing values in adjustment or more of the following variables had never been meas-
variables were imputed with last observation carried forward or ured or the test failed throughout the study: cytogenetics
next observation carried backward. (n=112), EQ-5D (n=101), blast count (n=60), platelet count
(n=1), and neutrophil count (n=2). The final cohort con-
sisted of 1,267 patients, unselected for any type of treat-
Results ment. In 162 patients the disease had progressed to high-
er-risk MDS or acute myeloid leukemia and 317 patients
Patients’ characteristics had died without progression. The median survival after
The EUMDS Registry contained data from 2,192 disease progression was 5.3 months [95% confidence
patients diagnosed between December 3, 2007 and March interval (95% CI); 3.2- 9.8 months]. Full details of the

Table 1. Baseline characteristics of the included patients from time of diagnosis and progression-free survival, stratified according to transfusion
status at the visit 3 landmark.
Total Hazard ratio Adjusted hazard Transfusion status at landmark
N. (%) (95% CI) ratio* (95% CI)
No Yes
Total 1267 (100.0) 751 (100.0) 516 (100.0)
Median age at diagnosis, 73.0 (18.0 - 95.0) 1.03 (1.02 - 1.04) 1.03 (1.02 - 1.04) 73.0 (18.0 - 91.0) 73.0 (21.0 - 95.0)
years (range)
Sex
Male 757 (59.7) 1 1 445 (59.3) 312 (60.5)
Female 510 (40.3) 0.84 (0.70 - 1.01) 0.76 (0.62 - 0.92) 306 (40.7) 204 (39.5)
WHO diagnosis:
RA 218 (17.2) 0.84 (0.64 - 1.10) 0.78 (0.59 - 1.03) 139 (18.5) 79 (15.3)
RARS 214 (16.9) 0.73 (0.56 - 0.96) 0.59 (0.45 - 0.78) 123 (16.4) 91 (17.6)
RCMD 492 (38.8) 1 1 296 (39.4) 196 (38.0)
RCMD-RS 86 (6.8) 1.03 (0.72 - 1.46) 0.91 (0.64 - 1.30) 47 (6.3) 39 (7.6)
RAEB-1 133 (10.5) 1.58 (1.20 - 2.07) 1.86 (1.41 - 2.46) 78 (10.4) 55 (10.7)
MDS-U 41 (3.2) 0.64 (0.34 - 1.22) 0.68 (0.36 - 1.29) 27 (3.6) 14 (2.7)
Deletion 5q 83 (6.6) 0.61 (0.40 - 0.92) 0.54 (0.35 - 0.83) 41 (5.5) 42 (8.1)
MDS Comorbidity Index
Low 782 (61.7) 1 1 482 (64.2) 300 (58.1)
Intermediate 411 (32.4) 1.24 (1.02 - 1.50) 1.08 (0.88 - 1.31) 232 (30.9) 179 (34.7)
High 71 (5.6) 1.55 (1.08 - 2.22) 1.30 (0.90 - 1.89) 35 (4.7) 36 (7.0)
Not known 3 (0.2) - - 2 (0.3) 1 (0.2)
Karnofsky status
80-100 881 (69.5) 1 1 543 (72.3) 338 (65.5)
50-70 210 (16.6) 1.72 (1.38 - 2.15) 1.40 (1.10 - 1.77) 93 (12.4) 117 (22.7)
10-40 10 (0.8) 2.04 (0.76 - 5.48) 1.89 (0.69 - 5.15) 3 (0.4) 7 (1.4)
Not known 166 (13.1) 1.08 (0.80 - 1.45) 0.99 (0.73 - 1.34) 112 (14.9) 54 (10.5)
Quality of life
Visual analog score, mean (SD) 70.5 (19.7) 0.99 (0.98 - 0.99) 0.99 (0.99 - 1.00) 73.1 (18.9) 66.8 (20.2)
IPSS category
Low 680 (53.7) 1 1 460 (61.3) 220 (42.6)
Intermediate 557 (44.0) 1.95 (1.62 - 2.34) 1.71 (1.39 - 2.11) 274 (36.5) 283 (54.8)
Cytogenetics not done 30 (2.4) 0.83 (0.43 - 1.62) 0.74 (0.38 - 1.45) 17 (2.3) 13 (2.5)
Revised IPSS category
Very low 386 (30.5) 1 1 310 (41.3) 76 (14.7)
Low 571 (45.1) 1.80 (1.41 - 2.29) 1.85 (1.45 - 2.37) 309 (41.1) 262 (50.8)
Intermediate 204 (16.1) 3.19 (2.41 - 4.22) 3.40 (2.55 - 4.52) 89 (11.9) 115 (22.3)
High 39 (3.1) 4.27 (2.72 - 6.71) 4.59 (2.91 - 7.22) 11 (1.5) 28 (5.4)
Very high 3 (0.2) 3.15 (0.78 - 12.82) 4.65 (1.13 - 19.15) 1 (0.1) 2 (0.4)
Not known 64 (5.1) 1.69 (1.07 - 2.68) 1.76 (1.11 - 2.80) 31 (4.1) 33 (6.4)
*Hazard ratio adjusted for all other variables in the table. 95% CI: 95% confidence Interval; WHO: World Health Organization; RA: refractory anemia; RARS: refractory anemia
with ring sideroblasts, RCMD: refractory cytopenia with multilineage dysplasia; RCMD-RS: refractory cytopenia with multilineage dysplasia & ring sideroblasts; RAEB: refractory
anemia with excess blasts; MDS-U: myelodysplastic syndrome, unclassifiable; MDS: myelodysplastic syndrome; SD: standard deviation; IPSS: International Prognostic Scoring
System.

634 haematologica | 2020; 105(3)

 Transfusion dose density in lower-risk MDS

exclusions are provided in the Online Supplementary Data. groups were 1.85 (95% CI: 1.24-2.76), and 3.79 (95% CI:
Table 1 and Online Supplementary Table S1 show the 2.65-5.42) relative to the non-transfused group. The
patients’ baseline demographics. For the landmark analy- recently revised International Working Group (IWG)
sis patients were defined as untransfused if they had never hematologic response criteria for patients with MDS
received a transfusion from diagnosis until the end of the refined RBCT burden by dividing patients into three cat-
study period (death or progression), or if they had egories (non-transfused patients, patients with a low
received transfusion only once (n=751). Patients were transfusion burden (0.75-2 units per month) and those
defined as transfused if they had received multiple trans- with a high transfusion burden (≥2 units per month).19
fusions (n=516) within the first year of follow-up (visit 3 = We therefore repeated the analysis, subdividing the
landmark). Regular transfusions were usually initiated patients into four groups: no transfusions, >0 to <0.75
during the first 6 months. Using visit 3 as the landmark (low transfusion burden), 0.75 to 1.75 (mid transfusion
ensured that the majority of patients who received more burden) and >1.75 (high transfusion burden). The results
than one transfusion were correctly identified. are shown in Figure 2D. The main effect occurred for
low dose densities, such that the outcomes of the mid
Distribution of transfusion dose density and high transfusion density groups were similar. The
The distribution of non-zero dose densities at the third low transfusion burden group of Figure 2D (density >0 -
visit (the landmark visit) is shown in Figure 1. Mean dose <0.75 units per month) is almost identical to the low bur-
density among those who had received a transfusion at 1 den group (density <0.89 units pwer month) of Figure 2B.
year of follow-up was 1.24 units per month, with a medi- MDS-related causes of death increased from 28% in the
an of 0.88 units per month (interquartile range, 0.31 – non-transfused group to 39% and 48% in the mid and
1.85). Dose densities of the transfused patients declined high transfusion burden groups, respectively (data not
on approach to the final recorded interval, if the patient shown).
died or progressed to higher-risk MDS during the last
interval (Online Supplementary Figure S1). This implies that Impact of individual prognostic factors
patients received fewer transfusions per month in the The univariate effect of various covariates on outcome
interval during which death occurred than in the preced- was investigated in order to discover the appropriate
ing intervals. Presumably, the treatment focus switches to functional form for the covariates (i.e., to discover
palliative care at home on the approach to death. Patients whether a linear or non-linear form was best) and to dis-
alive at the last recorded visit and with no signs of pro- cover appropriate ways of adjusting for confounding
gression did not show an increase of transfusion density covariates. Increasing RBCT dose density was associated
over time (Online Supplementary Figure S1). with inferior PFS (P<1x10-4). The functional form is
shown in Figure 3A. The effect of the dose density
Outcome of patients stratified according increased until a dose density of about 1 unit per month;
to transfusion status at the landmark 1 year thereafter, the effect was flat. Baseline age (as a continu-
after registration ous variable) was strongly associated with PFS (P<1x10-4)
The patients’ characteristics at the time of the landmark in univariate regression analyses, as were baseline MDS
visit 3 stratified according to transfusion status are shown diagnosis (P<1x10-4), quality of life measured by the EQ-
in Online Supplementary Table S2. One hundred forty-five 5D Index (P<1x10-4), country of origin (P=0.002), bone
subjects untransfused at visit 3 went on to have transfu- marrow blast count (P<1x10-4), number of cytopenias
sions after the landmark visit. Out of 516 transfused by (P<1x10-4), revised IPSS cytogenetic category (P<1x10-4),
the time of the landmark, 288 subjects were not reported hemoglobin concentration (P<1x10-4), neutrophil count
to have received any further transfusions, but of these (P<1x10-4) and platelet count (P<1x10-4). No difference in
288, 125 subjects did not have any further visits and
another 91 had only one additional visit. Of the 163 sub-
jects who had one or more additional visits (91+72,
respectively), 73 received treatment with ESA, 19 with
lenalidomide, ten with hypomethylating agents, two
with hydroxycarbamide, and three with iron chelators.
Unadjusted PFS stratified by transfusion status (trans-
fused n=516, untransfused n=751) at the third visit is pre-
sented in Figure 2A. The overall PFS of the untransfused
patients at visit 3 was significantly better (P<0.0001) than
that of the transfused patients.
Transfused patients were divided into those receiving
above (high density) or below (low density) the median
value (0.87 units per month) of non-zero dose densities.
Unadjusted PFS stratified by transfusion status and dose
density (untransfused n=751, low dose density n=258,
high dose density=258) at the third visit is presented in
Figure 2B. The overall PFS of the three groups of patients,
stratified according to the dose density at visit 3, was sig-
nificantly different (P<0.0001). We evaluated the time to
progression in the three groups of patients by censoring Figure 1. Distribution of dose densities of all transfused patients in the interval
preceding the 1-year landmark. Frequency: number of patients in each dose
density ranging from >0 to 0.2 units per month to >6 units per mont.
those who died before progression (Figure 2C). The haz-
ard ratios for the patients in the low and high density

haematologica | 2020; 105(3) 635

 L. de Swart et al.

PFS was detected by sex (P=0.1), but PFS in females was jects’ country of origin, all previously significant variables,
superior in the multivariate analyses. including the dose density, retained statistical significance,
with a dose density P-value of <10-4.
Progression-free survival using time-varying
covariates proportional hazards regression analysis Impact of therapeutic interventions on red blood cell
Variables used for adjustment at baseline included age at transfusion densities
diagnosis, sex, country of origin, number of cytopenias Treatment with ESA, lenalidomide and iron chelators
(and their corresponding blood counts), and number of may improve erythropoiesis and reduce the need for
units of blood received before registration. Time-varying RBCT. Reduction of the RBCT rate results in a gradual
variables measured longitudinally included: dose density, decrease of the subsequent RBCT dose densities in inter-
EQ-5D Index, components of the IPSS-R, and receipt of vals during the response period. We therefore investigated
ESA, iron chelators and lenalidomide. how many of the transfused patients had been treated
In multivariate analysis, not adjusting for the effects of with these interventions and calculated the average treat-
ESA, iron chelation and lenalidomide therapy, all variables ment duration and the number of patients with reduced
entered in the regression retained statistical significance. transfusion densities after starting the intervention. In our
The functional form of the dose density effect (P<10-4) is cohort of 1,267 patients, 679 received treatment with an
shown in Figure 3B. With a frailty term added for the sub- ESA and 151 had reduced transfusion densities in the first

A B

C D

Figure 2. Progression-free survival and risk of progression according to transfusion status at the landmark of visit 3 (1 year after registration). (A) Kaplan-Meier
plot of progression-free survival (PFS) of patients who did or did not receive transfusions by the landmark (visit 3). (B) Kaplan-Meier plot of PFS of patients who
received transfusions at a low density (<0.87 units/month) or at a high density (>0.87 units/month) by the landmark versus PFS of patients who did not receive
transfusions; (C) Kaplan-Meier plot of time to progression of patients surviving until progression subdivided according to transfusion burden or not as in panel B; (D)
Kaplan-Meier plot of PFS of patients receiving transfusions at densities according to the revised International Working Group criteria: low dose density: >0- <0.75
units per month; mid dose density: 0.75 - 1.75 units per month; high dose density >1.75 units per month.

636 haematologica | 2020; 105(3)

 Transfusion dose density in lower-risk MDS

visit after starting ESA treatment. Online Supplementary generally accepted. Several studies have addressed this
Figure S2 gives the individual dose density over time dur- question using various definitions of transfusion rate. The
ing ESA treatment of the 151 responding patients. Overall, initial publications describing the impact of RBCT on out-
100 patients received treatment with lenalidomide: of come in MDS compared RBCT-dependent patients with
these, 53 patients had a reduced transfusion density in the RBCT-independent patients, using RBCT dependency as a
first visit after starting lenalidomide treatment; Online time-dependent variable.1,21 These studies were based on
Supplementary Figure S3 shows the individual dose density various definitions of RBCT dependency,22,23 including a
over time during lenalidomide treatment of the 53 study using a rigid criterion, which implied a RBCT rate of
responding patients. Within our study group 186 patients at least 1 unit per month during a period of 2 months.24 In
received treatment with iron chelators and 75 patients had this last study, transfusion dependency occurred in a
a response leading to reduced transfusion densities in the minority of the patients (35% to 44%). The use of this
first interval after starting of iron chelation treatment
(Online Supplementary Figure S4). In contrast to the dose
densities over time during ESA and lenalidomide treat- A
ment, the longer-term dose densities during iron chelation
appeared to show a more stable pattern: subjects receiving
a certain level of blood transfusion dose density when
they first received iron chelation appeared to maintain
that level of dose density. The decline of the dose density
was less pronounced, but this might be a reflection of the
longer transfusion period before starting chelation treat-
ment when compared with the other two interventions.
The observed patterns of dose density trajectories sug-
gest that receiving ESA, lenalidomide or iron chelation
therapy modulates the dose density and we, therefore,
included these variables in the regression model. This
analysis resulted in an effect for the dose density similar to
that of the previous analyses (Figure 3B), with a P-value of
<0.0001. Indeed all variables entered in the regression B
retained statistical significance, except for platelet count
(P=0.47) and neutrophil count (P=0.24). However, the
dose density effect continued to increase beyond 1 unit
per month after correction for the three interventions
(ESA, iron chelation and lenalidomide) up until a dose of 6
units per month (Figure 3C).
Some patients received more than one intervention
simultaneously, including 25 patients who received chela-
tion and lenalidomide and 88 patients who received ESA
and chelation. However, no additional impact could be
detected over and above the impact of the two individual
interventions.

Discussion C

This large prospective, observational study confirmed

the reported association of transfusion dose density with
reduced PFS in patients with lower-risk MDS.20 More sur-
prisingly, we showed in this study that this negative asso-
ciation already occurred at a low transfusion rate. In addi-
tion, we showed that the risk of progression increased
both in the low and high transfusion burden groups when
compared to the non-transfused patients. We were even
able to show that the deleterious effect of transfusions
occurred at a very low transfusion burden (<0.75 units per
month or <3 units per 16 weeks as defined by the revised
IWG criteria), when the patients were subdivided accord-
ing to the revised IWG hematological response criteria.19
These patients with a very low transfusion burden are
Figure 3. Influence of dose density on progression-free survival. (A) Dose density
considered as untransfused patients using the revised IWG
response criteria.19 effect on progression-free survival (PFS) in an univariate analysis. (B) Dose den-
The main focus of our study was to analyze the associ- sity effect on PFS in a multivariate regression model unadjusted for the three
treatment variables. (C) Dose density effect on PFS in a multivariate regression
model adjusted for treatment with either erythropoiesis-stimulating agent, Iron
ation of transfusion rate with outcome, assuming that reg-
ularly transfused patients may be exposed to the postulat- chelation or lenalidomide.
ed toxicity of RBCT at a lower transfusion burden than

haematologica | 2020; 105(3) 637

 L. de Swart et al.

definition implies that patients regularly receiving fewer Many patients showed a (temporary) decrease of the
than 3 units per 16 weeks are defined as RBCT-indepen- RBCT dose density, reflecting response to ESA,27 lenalido-
dent, but these patients might also be subject to the dele- mide,28 and/or iron chelators12 in 22%, 53% and 40% of
terious association with RBCT. In addition, patients may the treated patients, respectively. The observed patterns of
respond to therapeutic interventions, such as ESA, dose density trajectories suggest that receipt of ESA,
lenalidomide or iron chelators and become RBCT-inde- lenalidomide and/or iron chelation modulates the dose
pendent again. The conclusion was that the severity of density and we therefore included these variables as con-
anemia was the leading cause of impaired survival rather founding variables in the regression model. This analysis
than RBCT dependency.24 However, the definition of showed that the impact of the dose density remained sim-
severe anemia (<9 g/dL in males and <8 g/dL in females) ilar to that in the previous analyses, but in contrast to the
implies that the majority of these patients were regularly previous analyses there is some evidence that the dose
transfused, as confirmed in the study.24 This study also density effect continued to increase beyond 2 units per
showed that the transfusion rate was significantly associ- month after correction for the three interventions.
ated with an increased risk of cardiac complications. The Red blood cells are usually transfused after a certain
risk of cardiac complications was significantly higher in period of storage, but the survival of stored red blood cells
patients with a RBCT intensity of >3 units per month depends on this period.29,30 Transfusion of stored red cells
compared to that in patients transfused with <1 unit per leads to pro-inflammatory reactions, associated with a
month.24 In an open forum discussion RBCT dependency higher risk of infection and increased levels of circulating
was even defined much higher, at 2 units per month in a iron and, in particular, NTBI species, which enhance bac-
3-month interval.25 In a Spanish study of 191 transfused terial growth in vitro.31,32 Infusion of autologous red blood
patients with MDS, the interval between each transfusion cells from healthy volunteers after prolonging storage up
was used to calculate the transfusion intensity.26 It was to 6 weeks resulted in increased extravascular hemolysis,
concluded that high transfusion intensity was associated decreased red cell survival, elevated NTBI and ferritin lev-
with decreased survival and increased risk of development els in units transfused after 6 weeks compared to units
of acute myeloid leukemia, in concordance with our transfused after shorter shortage.33 Excess toxic iron
study. Interestingly, the cumulative transfusion burden species, including NTBI and especially its component
was not a prognostic factor when the transfusion intensity LPI,34 catalyze the cellular generation of reactive oxygen
was included in the model.26 species. Oxidative stress may lead to pro-inflammatory
The traditional evaluations of the prognostic impact of responses and to oxidation of lipids, proteins and DNA
factors influencing outcome have used standard time-to- causing cell and tissue damage.35,36 Elevated NTBI levels
event methods based on variables at diagnosis; however, after a single unit of RBC stored for 6 weeks normalize
many variables in MDS may change over time. This within 24 hours.37 However, in multi-transfused patients
aspect can be addressed by using proportional hazards (cumulative number of units ≥10) with MDS, NTBI and
regression with time-varying covariates. The EUMDS LPI remained elevated until the next transfusion.17
Registry is collecting its observational data at registration In conclusion, the negative association of transfusions
of each new patient (within 100 days after diagnosis) and on PFS already occurs at low RBCT dose densities below
follow-up data at 6-monthly intervals. This practice leads 3 units per 16 weeks. This indicates that the RBCT
to regular visit intervals of 6 months. For many patients in dependency in patients transfused at relatively low rates,
this dataset, the value of the recorded transfusion rate var- who are usually considered as untransfused patients, may
ied strongly over time, as shown in the Online be considered as an indicator of poor prognosis for PFS.
Supplementary Files. We therefore calculated the RBCT rate This poor prognosis in transfusion-dependent patients
at each reported visit during all preceding visit intervals might be the result of direct toxicity of iron radicals result-
between the date of the first RBCT and the date of the last ing from the RBCT or the result of concomitant disease
visit, leading to a “smoothed” variable, defined as dose progression, including hematopoietic impairment. Data
density. This reflects an average rate of receiving transfu- from our group provide support for the direct toxicity of
sions during the whole observation period with transfu- RBCT density on outcome, because patients had a better
sions. The relatively low number of red cell units trans- outcome if treated with chelators, which remove toxic
fused per month can be explained by the remarkable vari- iron radicals effectively. Future studies, including interven-
ation of the transfusion rate over time, even when using tional studies, are needed to confirm our observations,
interval visit reports of 6 months’ duration. which may lead to adaptions of the current recommenda-
Baseline age, bone marrow percentage category, num- tions.
ber of cytopenias, and the EQ-5D Index retained their sig-
nificant prognostic impact in the proportional hazards Acknowledgments
regression with time-varying explanatory variables. The The authors and members of the steering committee of the
non-linear component of the dose density effect was also EUMDS Registry would like to thank all local investigators and
retained (P<1x10-4). The unfavorable effect of the dose operational team members for their contribution.
density increased until a dose density of about 2 units per
month and leveled off thereafter. A similar form and effect Funding
was observed when using the cumulative dose of RBCT The work of the EUMDS Registry is supported by an educa-
units over time in an identical multivariate regression tional grant from Novartis Pharmacy B.V. Oncology Europe, and
model with the same variables (data not shown). The nega- Amgen Limited. This work is part of the MDS-RIGHT activities,
tive impact of the cumulative RBCT dose started already which has received funding from the European Union’s Horizon
at the time of administering the first RBCT and did not 2020 research and innovation programme under grant agreement
increase any further beyond 30 units received (data not n. 634789 - “Providing the right care to the right patient with
shown). MyeloDysplastic Syndrome at the right time”.

638 haematologica | 2020; 105(3)

 Transfusion dose density in lower-risk MDS

(6):463-467. Transfusion intensity, not the cumulative

References 14. De Souza GF, Ribeiro HL, Jr., De Sousa JC, et red blood cell transfusion burden, deter-
1. Cazzola M, Malcovati L. Myelodysplastic al. HFE gene mutation and oxidative dam- mines the prognosis of patients with
syndromes--coping with ineffective age biomarkers in patients with myelodys- myelodysplastic syndrome on chronic trans-
hematopoiesis. N Engl J Med. 2005;352 plastic syndromes and its relation to transfu- fusion support. Am J Hematol. 2011;86
(6):536-538. sional iron overload: an observational cross- (3):245-250.
2. Malcovati L, Porta MG, Pascutto C, et al. sectional study. BMJ Open. 2015;5(4): 27. Hellstrom-Lindberg E, Negrin R, Stein R, et
Prognostic factors and life expectancy in e006048. al. Erythroid response to treatment with G-
myelodysplastic syndromes classified 15. Saigo K, Takenokuchi M, Hiramatsu Y, et al. CSF plus erythropoietin for the anaemia of
according to WHO criteria: a basis for clini- Oxidative stress levels in myelodysplastic patients with myelodysplastic syndromes:
cal decision making. J Clin Oncol. 2005;23 syndrome patients: their relationship to proposal for a predictive model. Br J
(30):7594-7603. serum ferritin and haemoglobin values. J Int Haematol. 1997;99(2):344-351.
3. Greenberg P, Cox C, LeBeau MM, et al. Med Res. 2011;39(5):1941-1945. 28. Fenaux P, Giagounidis A, Selleslag D, et al. A
International scoring system for evaluating 16. Bulycheva E, Rauner M, Medyouf H, et al. randomized phase 3 study of lenalidomide
prognosis in myelodysplastic syndromes. Myelodysplasia is in the niche: novel con- versus placebo in RBC transfusion-depen-
Blood. 1997;89(6):2079-2088. cepts and emerging therapies. Leukemia. dent patients with low-/intermediate-1-risk
4. de Swart L, Smith A, Johnston TW, et al. 2015;29(2):259-268. myelodysplastic syndromes with del5q.
Validation of the revised international prog- 17. de Swart L, Reiniers C, Bagguley T, et al. Blood. 2011;118(14):3765-3776.
nostic scoring system (IPSS-R) in patients Labile plasma iron levels predict survival in 29. Luten M, Roerdinkholder-Stoelwinder B,
with lower-risk myelodysplastic syndromes: patients with lower-risk myelodysplastic Schaap NP, de Grip WJ, Bos HJ, Bosman GJ.
a report from the prospective European syndromes. Haematologica. 2018;103(1):69- Survival of red blood cells after transfusion:
LeukaemiaNet MDS (EUMDS) registry. Br J 79. a comparison between red cells concentrates
Haematol. 2015;170(3):372-383. 18. Vardiman JW, Harris NL, Brunning RD. The of different storage periods. Transfusion.
5. Zipperer E, Post JG, Herkert M, et al. Serum World Health Organization (WHO) classifi- 2008;48(7):1478-1485.
hepcidin measured with an improved ELISA cation of the myeloid neoplasms. Blood. 30. van de Watering L, Lorinser J, Versteegh M,
correlates with parameters of iron metabo- 2002;100(7):2292-2302. Westendord R, Brand A. Effects of storage
lism in patients with myelodysplastic syn- 19. Platzbecker U, Fenaux P, Ades L, et al. time of red blood cell transfusions on the
drome. Ann Hematol. 2013;92(12):1617- Proposals for revised IWG 2018 hematolog- prognosis of coronary artery bypass graft
1623. ical response criteria in patients with MDS patients. Transfusion. 2006;46(10):1712-
6. Santini V, Girelli D, Sanna A, et al. Hepcidin included in clinical trials. Blood. 1718.
levels and their determinants in different 2018;133(10):1020-1030. 31. Hod EA, Zhang N, Sokol SA, et al.
types of myelodysplastic syndromes. PLoS 20. Goldberg SL, Chen E, Corral M, et al. Transfusion of red blood cells after pro-
One. 2011;6(8):e23109. Incidence and clinical complications of longed storage produces harmful effects that
7. Cuijpers ML, Raymakers RA, Mackenzie myelodysplastic syndromes among United are mediated by iron and inflammation.
MA, de Witte TJ, Swinkels DW. Recent States Medicare beneficiaries. J Clin Oncol. Blood. 2010;115(21):4284-4292.
advances in the understanding of iron over- 2010;28(17):2847-2852. 32. McQuilten ZK, French CJ, Nichol A, Higgins
load in sideroblastic myelodysplastic syn- 21. Hiwase DK, Singhal D, Strupp C, et al. A, Cooper DJ. Effect of age of red cells for
drome. Br J Haematol. 2010;149(3):322-333. Dynamic assessment of RBC-transfusion transfusion on patient outcomes: a system-
8. Shenoy N, Vallumsetla N, Rachmilewitz E, dependency improves the prognostic value atic review and meta-analysis. Transfus
Verma A, Ginzburg Y. Impact of iron over- of the revised-IPSS in MDS patients. Am J Med Rev. 2018;32(2):77-88.
load and potential benefit from iron chela- Hematol. 2017;92(6):508-514. 33. Rapido F, Brittenham GM, Bandyopadhyay
tion in low-risk myelodysplastic syndrome. 22. Malcovati L, Germing U, Kuendgen A, et al. S, et al. Prolonged red cell storage before
Blood. 2014;124(6):873-881. Time-dependent prognostic scoring system transfusion increases extravascular hemoly-
9. Ambaglio I, Malcovati L, Papaemmanuil E, for predicting survival and leukemic evolu- sis. J Clin Invest. 2017;127(1):375-382.
et al. Inappropriately low hepcidin levels in tion in myelodysplastic syndromes. J Clin 34. de Swart L, Hendriks JC, van der Vorm LN,
patients with myelodysplastic syndrome Oncol. 2007;25(23):3503-3510. et al. Second international round robin for
carrying a somatic mutation of SF3B1. 23. Waszczuk-Gajda A, Madry K, Machowicz the quantification of serum non-transfer-
Haematologica. 2013;98(3):420-423. R, et al. Red blood cell transfusion depend- rin-bound iron and labile plasma iron in
10. Rund D, Rachmilewitz E. Beta-thalassemia. ency and hyperferritinemia are associated patients with iron-overload disorders.
N Engl J Med. 2005;353(11):1135-1146. with impaired survival in patients diagnosed Haematologica. 2016;101(1):38-45.
11. Gattermann N, Rachmilewitz EA. Iron over- with myelodysplastic syndromes: results 35. Rachmilewitz EA, Weizer-Stern O,
load in MDS-pathophysiology, diagnosis, from the first Polish MDS-PALG registry. Adamsky K, et al. Role of iron in inducing
and complications. Ann Hematol. 2011;90 Adv Clin Exp Med. 2016;25(4):633-641. oxidative stress in thalassemia: can it be pre-
(1):1-10. 24. Malcovati L, Della Porta MG, Strupp C, et al. vented by inhibition of absorption and by
12. Gattermann N, Finelli C, Della Porta M, et Impact of the degree of anemia on the out- antioxidants? Ann N Y Acad Sci. 2005;1054:
al. Hematologic responses to deferasirox come of patients with myelodysplastic syn- 118-123.
therapy in transfusion-dependent patients drome and its integration into the WHO 36. Hershko C, Link G, Cabantchik I.
with myelodysplastic syndromes. classification-based Prognostic Scoring Pathophysiology of iron overload. Ann N Y
Haematologica. 2012;97(9):1364-1371. System (WPSS). Haematologica. 2011;96 Acad Sci. 1998;850:191-201.
13. Ghoti H, Amer J, Winder A, Rachmilewitz (10):1433-1440. 37. Hod EA, Brittenham GM, Billote GB, et al.
E, Fibach E. Oxidative stress in red blood 25. Gale RP, Barosi G, Barbui T, et al. What are Transfusion of human volunteers with older,
cells, platelets and polymorphonuclear RBC-transfusion-dependence and -indepen- stored red blood cells produces extravascular
leukocytes from patients with myelodys- dence? Leuk Res. 2011;35(1):8-11. hemolysis and circulating non-transferrin-
plastic syndrome. Eur J Haematol. 2007;79 26. Pereira A, Nomdedeu M, Aguilar JL, et al. bound iron. Blood. 2011;118(25):6675-6682.

haematologica | 2020; 105(3) 639