Article
Reference
High prognostic impact of flow cytometric minimal residual disease
detection in acute myeloid leukemia: data from the HOVON/SAKK
AML 42A study
TERWIJN, Monique, et al.
Abstract
Half the patients with acute myeloid leukemia (AML) who achieve complete remission (CR),
ultimately relapse. Residual treatment-surviving leukemia is considered responsible for the
outgrowth of AML. In many retrospective studies, detection of minimal residual disease (MRD)
has been shown to enable identification of these poor-outcome patients by showing its
independent prognostic impact. Most studies focus on molecular markers or analyze data in
retrospect. This study establishes the value of immunophenotypically assessed MRD in the
context of a multicenter clinical trial in adult AML with sample collection and analysis
performed in a few specialized centers.
TERWIJN, Monique, et al. High prognostic impact of flow cytometric minimal residual disease
detection in acute myeloid leukemia: data from the HOVON/SAKK AML 42A study. Journal of
clinical oncology, 2013, vol. 31, no. 31, p. 3889-97
DOI : 10.1200/JCO.2012.45.9628
PMID : 24062400
Available at:
http://archive-ouverte.unige.ch/unige:37145
Disclaimer: layout of this document may differ from the published version.
1 / 1
High Prognostic Impact of Flow Cytometric Minimal
Residual Disease Detection in Acute Myeloid Leukemia:
Data From the HOVON/SAKK AML 42A Study
Monique Terwijn, Wim L.J. van Putten, Ange`le Kelder, Vincent H.J. van der Velden, Rik A. Brooimans,
Thomas Pabst, Johan Maertens, Nancy Boeckx, Georgine E. de Greef, Peter J.M. Valk, Frank W.M.B. Preijers,
Peter C. Huijgens, Angelika M. Dra¨ger, Urs Schanz, Mojca Jongen-Lavrecic, Bart J. Biemond, Jakob R. Passweg,
Michel van Gelder, Pierre Wijermans, Carlos Graux, Mario Bargetzi, Marie-Cecile Legdeur, Jurgen Kuball,
Okke de Weerdt, Yves Chalandon, Urs Hess, Leo F. Verdonck, Jan W. Gratama, Yvonne J.M. Oussoren,
Willemijn J. Scholten, Jennita Slomp, Alexander N. Snel, Marie-Christiane Vekemans, Bob Lo¨wenberg,
Gert J. Ossenkoppele, and Gerrit J. Schuurhuis
See accompanying editorial on page 3857
Author affiliations appear at the end of
this article.
Published online ahead of print at
www.jco.org on September 23, 2013.
Supported in part by Grant No. KWF
2006-3695 from the Dutch Cancer
Society.
Authors’ disclosures of potential con-
flicts of interest and author contribu-
tions are found at the end of this
article.
Corresponding author: Gerrit J.
Schuurhuis, PhD, VU University Medical
Center, Department of Hematology,
CCA-4.28, PO Box 7057, 1007 MB
Amsterdam, the Netherlands; e-mail:
gj.schuurhuis@vumc.nl.
© 2013 by American Society of Clinical
Oncology
0732-183X/13/3131w-3889w/$20.00
DOI: 10.1200/JCO.2012.45.9628
ABSTRACT
Purpose
Half the patients with acute myeloid leukemia (AML) who achieve complete remission (CR),
ultimately relapse. Residual treatment-surviving leukemia is considered responsible for the
outgrowth of AML. In many retrospective studies, detection of minimal residual disease (MRD)
has been shown to enable identification of these poor-outcome patients by showing its
independent prognostic impact. Most studies focus on molecular markers or analyze data in
retrospect. This study establishes the value of immunophenotypically assessed MRD in the
context of a multicenter clinical trial in adult AML with sample collection and analysis performed
in a few specialized centers.
Patients and Methods
In adults (younger than age 60 years) with AML enrolled onto the Dutch-Belgian Hemato-Oncology
Cooperative Group/Swiss Group for Clinical Cancer Research Acute Myeloid Leukemia 42A study,
MRD was evaluated in bone marrow samples in CR (164 after induction cycle 1, 183 after cycle
2, 124 after consolidation therapy).
Results
After all courses of therapy, low MRD values distinguished patients with relatively favorable
outcome from those with high relapse rate and adverse relapse-free and overall survival. In the
whole patient group and in the subgroup with intermediate-risk cytogenetics, MRD was an
independent prognostic factor. Multivariate analysis after cycle 2, when decisions about consoli-
dation treatment have to be made, confirmed that high MRD values (⬎ 0.1% of WBC) were
associated with a higher risk of relapse after adjustment for consolidation treatment time-
dependent covariate risk score and early or later CR.
Conclusion
In future treatment studies, risk stratification should be based not only on risk estimation assessed
at diagnosis but also on MRD as a therapy-dependent prognostic factor.
J Clin Oncol 31:3889-3897. © 2013 by American Society of Clinical Oncology
INTRODUCTION
Acute myeloid leukemia (AML) is characterized by
an abnormal proliferation of myeloid progenitor
cells and subsequent bone marrow (BM) failure.
Despite high remission rates after intensive chemo-
therapy, 5-year survival is only approximately 30%
to 40%. Apart from increasing complete remission
(CR) rates, an important goal for treatment, guided
by prognostic factors at diagnosis, is to tune clinical
management in the postremission phase. Currently,
the most important prognostic factors at diagnosis
encompass cytogenetics and molecular abnorma-
lities.
1-4
Although of utmost importance in risk
stratification, treatment outcome for specifically de-
fined risk groups is still highly variable, especially in
intermediate-risk AML. Thus, there is a need for
additional prognostic factors, which may include
treatment- and response-related factors. In several
correlative studies, minimal residual disease (MRD)
JOURNAL OF CLINICAL ONCOLOGY
ORIGINAL REPORT
VOLUME 31 䡠 NUMBER 31 䡠 NOVEMBER 1 2013
© 2013 by American Society of Clinical Oncology
3889
Zurich) on October 29, 2013 from 129.195.0.205
Information downloaded from jco.ascopubs.org and provided by at SWISS CONSORTIUM (Hauptbibliothek Universitat
Copyright © 2013 American Society of Clinical Oncology. All rights reserved.
Zurich) on October 29, 2013 from 129.195.0.205
Information downloaded from jco.ascopubs.org and provided by at SWISS CONSORTIUM (Hauptbibliothek Universitat
Copyright © 2013 American Society of Clinical Oncology. All rights reserved.
Zurich) on October 29, 2013 from 129.195.0.205
Information downloaded from jco.ascopubs.org and provided by at SWISS CONSORTIUM (Hauptbibliothek Universitat
Copyright © 2013 American Society of Clinical Oncology. All rights reserved.
Zurich) on October 29, 2013 from 129.195.0.205
Information downloaded from jco.ascopubs.org and provided by at SWISS CONSORTIUM (Hauptbibliothek Universitat
Copyright © 2013 American Society of Clinical Oncology. All rights reserved.
Zurich) on October 29, 2013 from 129.195.0.205
Information downloaded from jco.ascopubs.org and provided by at SWISS CONSORTIUM (Hauptbibliothek Universitat
Copyright © 2013 American Society of Clinical Oncology. All rights reserved.
Zurich) on October 29, 2013 from 129.195.0.205
Information downloaded from jco.ascopubs.org and provided by at SWISS CONSORTIUM (Hauptbibliothek Universitat
Copyright © 2013 American Society of Clinical Oncology. All rights reserved.
Zurich) on October 29, 2013 from 129.195.0.205
Information downloaded from jco.ascopubs.org and provided by at SWISS CONSORTIUM (Hauptbibliothek Universitat
Copyright © 2013 American Society of Clinical Oncology. All rights reserved.
Zurich) on October 29, 2013 from 129.195.0.205
Information downloaded from jco.ascopubs.org and provided by at SWISS CONSORTIUM (Hauptbibliothek Universitat
Copyright © 2013 American Society of Clinical Oncology. All rights reserved.
Zurich) on October 29, 2013 from 129.195.0.205
Information downloaded from jco.ascopubs.org and provided by at SWISS CONSORTIUM (Hauptbibliothek Universitat
Copyright © 2013 American Society of Clinical Oncology. All rights reserved.
Zurich) on October 29, 2013 from 129.195.0.205
Information downloaded from jco.ascopubs.org and provided by at SWISS CONSORTIUM (Hauptbibliothek Universitat
Copyright © 2013 American Society of Clinical Oncology. All rights reserved.
has been convincingly shown to provide such additional prognostic
information.
5-15
Of interest, in the study by Rubnitz et al,
16
MRD
remained a prognostic factor in patients in whom treatment was
intensified on the basis of MRD positivity. MRD is defined as leukemic
cells persisting after chemotherapy below the sensitivity (detection
limit) of routine morphology. The most widely used techniques to
assess MRD in AML use molecular or immunophenotypic aberran-
cies. For the immunophenotypic identification of MRD, aberrantly
expressed markers are combined with normal myeloid antigens and,
when possible, progenitor markers, resulting in a so-called leukemia-
associated phenotype (LAP), which must be established at diagnosis.
Because the LAP is not present on normal cells (or is present at
relatively low frequencies), remission BM can be analyzed for LAP-
positive MRD cells with sensitivities ranging from 10
⫺3
to 10
⫺5
(one
leukemic cell in 1,000 to 100,000 WBCs).
7,9,12,17,18
It is important to
realize that these studies in patients with AML were performed retro-
spectively, mostly in a single-institute setting, thereby introducing
well-known potential bias. By identifying cut point values on the order
of 0.01% to 0.1%, it has been possible to identify patients markedly
differing in prognosis. In our retrospective study, we showed similar
results with cut points usable in a range of 0.06% to 1%.
10
In this study,
we set out to prospectively validate the previously defined cut points in
a multicenter international clinical trial (Dutch-Belgian Hemato-
Oncology Cooperative Group/Swiss Group for Clinical Cancer Re-
search Acute Myeloid Leukemia [HOVON/SAKK AML] 42A). To
that end, we determined MRD percentages in a setting in which MRD
assessment was performed without prior knowledge of clinical man-
agement, diagnostic features, or outcome. This study differs from
other studies
6,7,10,12,19
since the patients were enrolled onto a large
multicenter clinical study with preplanned sample collection and
MRD analysis in a few specialized centers according to common
protocols. The results show that by using methods established earlier,
the previously defined cut points
10
are highly predictive for clini-
cal outcome.
PATIENTS AND METHODS
The Data Supplement provides more detailed information.
Patients and Treatment
A total of 517 patients between the ages of 18 and 60 years were included
in this study (Fig 1). Patients were randomly assigned to receive granulocyte
colony-stimulating factor (G-CSF; 5
g/kg) or no G-CSF during induction
treatment. Data for these two groups were pooled since clinically there is no
difference in survival.
20
In agreement with this, no significant differences were
found between groups in MRD percentage after all therapy cycles. Patients
were assigned to three risk groups on the basis of the following criteria: (1)
good-risk patients included those positive for t(8;21) with WBC ⱕ 20 ⫻ 10
9
/L,
(2) those with inv(16) or t(16;16) and (3) those without a monosomal karyo-
type but with mutated CEBP
␣
and those with mutated NPM1/FLT3 wild-type
in CR after the first induction cycle. Poor-risk patients were defined as having
non– core-binding factor leukemia with a monosomal karyotype, being posi-
tive for EVI1, or having 3q26 abnormalities. The remaining patients were
classified as having intermediate-risk disease. Patient and treatment character-
istics are provided in Table 1.
Sampling and Logistics
Thirty-one centers participated in collection of patient samples as a part
of the MRD side study of the HOVON/SAKK AML 42A clinical study. LAP
assessment and MRD analysis were performed in four centers. A summary of
Patients included
(N = 517)
Patients available for analysis
(n = 389)
)821 = n( dedulcxE
)76 = n( PAL oN
Samples missed at diagnosis (n = 42)
Not evaluable (unfit material)* (n = 12)
Gave no consent (n = 3)
Ineligible for survival analysis (n = 4)
In CR directly after cycle 1
(n = 208)
Received 1st cycle
of chemotherapy
(n = 389)
Received 2nd cycle
of chemotherapy
(n = 331)
Received 3rd cycle
of chemotherapy
(n = 84)
Received transplantation
in CR1
(n = 186)
In CR (n = 269)
After cycle 1 (n = 194)
After cycle 2 (n = 75)
In CR (n = 83)
After cycle 1 (n = 56)
After cycle 2 (n = 20)
After transplantation (n = 7)
In CR (n = 186)
After cycle 1 (n = 134)
After cycle 2 (n = 50)
After transplantation (n = 2)
Samples sent for MRD analysis†
(n = 164)
Samples sent for MRD analysis†
(n = 183)
Samples sent for MRD analysis†
(n = 40)
Samples sent for MRD analysis†
(n = 84)
Fig 1. Diagram of patients included the HOVON/SAKK AML 42A study. Of 517 patients included, 389 patients showed one or more leukemia-associated phenotypes
(LAPs) at diagnosis and were suitable for the monitoring of minimal residual disease (MRD) in remission bone marrow. Bone marrow samples from 164 patients were
available for MRD analysis after the first cycle of chemotherapy, 183 samples were available after the second cycle, 40 samples were available after the third cycle,
and 84 samples were available after transplantation. (*) Due to dry tap and no blasts in peripheral blood or poor quality material with mainly dead cells. (†) Drop-off could
be only partly explained with death or relapse before the next cycle could be given or within 3 months after consolidation treatment (cycle 1, n ⫽ 0; cycle 2, n ⫽ 3;
cycle 3, n ⫽ 13; transplantation, n ⫽ 11). Other missing samples were not received. CR, complete response; CR1, first CR.
Terwijn et al
3890
© 2013 by American Society of Clinical Oncology
J
OURNAL OF CLINICAL ONCOLOGY
the logistics of sampling and MRD analysis is provided in the Data Supple-
ment. BM samples were collected at diagnosis and at follow-up after every
cycle of chemotherapy. LAP assessment and MRD analysis were performed in
a setting in which the laboratories had no access to patients’ clinical data until
final MRD data were sent to the statistician. In addition, clinicians had no
access to the MRD status of their patients. The data were included only for
patients with a morphologic CR. Figure 1 shows the details of sam-
pling outcome.
LAP Assessment
LAP assessment was performed in collaboration with the Dutch-Belgian
MRD flow cytometry taskforce and was done in a two-step procedure. As a
first step, a standard screening panel was designed to assess the immunophe-
notype of blasts, identified as dim expression of CD45 with low sideward
scatter properties (Data Supplement). The second step consisted of the valida-
tion of the composed LAP by showing its actual presence on the leukemic cells,
which should be on at least 10% of the blast population. The Data Supplement
gives an overview of different LAPs identified in the study. The Data Supple-
ment also shows LAP types as detected at diagnosis as well as LAPs actually
used in follow-up for MRD assessment. For refractory anemia with excess
blasts in transformation, only LAPs that covered the blast fraction were de-
fined; here, no LAPs present on mature cells were used. Information on clones
and commercial sources of all monoclonal antibodies used is provided in the
Data Supplement.
MRD Detection
MRD analysis was performed as previously described.
10
Analysis of LAP-
positive cells included multiple backgating steps to ensure that, compared with
diagnosis, the LAP-positive cells show fairly identical positions in forward
scatter channel/side scatter channel and CD45 expression. By using this
method, LAP-positive cell populations could be distinguished from back-
ground expression in the gate. MRD percentage was defined as the percentage
of LAP-positive cells within the WBC compartment multiplied by the correc-
tion factor: 100%/percentage of LAP-positive blasts at diagnosis. When there
was a considerable amount of background expression in the gate, the correc-
tion factor
10
was set to 1. In addition, calculations of MRD percentage have
also been performed by using uncorrected LAP-positive frequencies, both as
percentage of WBC and as log reduction of LAP-positive cells (in both cases,
LAP-positive events are relative to WBC count). MRD was also defined by
quantitative reverse transcriptase polymerase chain reaction (qRT-PCR)–
based log reduction. Total RNA was extracted, and complementary DNA
(cDNA) was synthesized from 500 ng of RNA by using random hexamer
priming, essentially as described.
21
Statistical Analysis
Separate analyses were performed for three landmarks: sample in CR after
cycle 1, after cycle 2, and after consolidation treatment. Primary end point for all
analyses was relapse with censoring at death in first CR (CR1). Secondary end
points were relapse-free survival (RFS), in which death in CR1 was included as a
competing risk event, and overall survival (OS). In each landmark analysis, time
was measured from the date of sampling. RFS, OS, and relapse incidence curves
22
were calculated according to Kaplan-Meier. In addition, competing risks actuarial
estimates of relapse and death in CR1 at 4 years were estimated by cumulative
incidence functions
23
and are presented in Table 2.
RESULTS
Regression Analysis for Corrected and Uncorrected
MRD Percentage and LAP-Positive Log Reduction
MRD percentages were assessed by including a correction factor
as described in Patients and Methods. In addition, we investigated the
prognostic impact of the percentage of LAP-positive cells without a
correction factor and the log reduction of LAP-positive cells, an ap-
proach previously used by Kern et al.
11
For each of the three
covariates—log-transformed percentage of MRD,
10
log-transformed
percentage of LAP-positive cells, and LAP-positive cell log reduc-
tion—Cox regression analysis after landmark cycle 2 with the end
point of relapse was done with adjustment for AML risk and early or
late CR. All three covariates showed a highly significant association
with risk of relapse: hazard ratio (HR), 1.49 (P ⫽ .007); HR, 1.50 (P ⫽
.015); and HR, 0.66 (P ⫽ .009), respectively. Note that a high log
reduction of LAP-positive cells corresponds with a low percentage of
LAP-positive cells. After landmark consolidation, similar results were
found: HR, 3.2 (P ⫽ 3.0 ⫻ 10
⫺8
); HR, 3.8 (P ⫽ 1.6 ⫻ 10
⫺6
); and HR,
0.38 (P ⫽ 7 ⫻ 10
⫺6
), respectively.
To validate our previous results (described in the following para-
graphs), further analyses were performed, mainly with corrected
MRD percentages.
10
The highly significant association seen between
MRD percentage as a continuous covariate and the risk of relapse
allows searching for optimal cut points established in a wide range
Table 1. Patient and Treatment Characteristics
Characteristic
No. of
Patients at
Diagnosis %
No. of
Patients
With MRD
ⱕ 0.1%
After
Cycle 2 %
No. of
Patients
With MRD
⬎ 0.1%
After
Cycle 2 %
Total 241
ⴱ
141 42
Sex
Male 122 51 73 24
Female 119 49 68 18
Age, years
Median 48 48 43
Range 18-60 18-60 21-58
ⱕ 40 71 29 44 31 17 40
⬎ 40 170 71 97 69 25 60
WBC at diagnosis (⫻10
9
/L)
ⱕ 20 136 56 87 62 18 43
20-100 69 29 42 30 10 24
⬎ 100 36 15 12 9 14 33
AML type
De novo AML 203 84 121 86 34 81
Secondary AML 21 9 8 6 7 17
RAEB 6 2 5 4 1 2
RAEB-t 11 5 7 5 0 0
Consolidation treatment
None 32 13 15 11 6 14
Cycle 3 52 22 26 18 14 33
Autologous SCT 65 27 44 31 10 24
Allogeneic SCT 92 38 56 40 12 29
Risk group
Good 64 27 38 27 14 33
Intermediate 143 59 88 62 19 45
Poor 34 14 15 11 9 21
CR achieved
After cycle 1 181 75 114 81 23 55
After cycle 2 60 25 27 19 19 45
G-CSF
Did not receive G-CSF 115 48 64 45 19 45
Received G-CSF 126 52 77 55 23 55
Abbreviations: AML, acute myeloid leukemia; CR, complete response;
G-CSF, granulocyte colony-stimulating factor; MRD, minimal residual disease;
RAEB, refractory anemia with excess blasts; RAEB-t, RAEB in transformation;
SCT, stem-cell transplantation.
ⴱ
Total No. of patients available for MRD analysis in whom at least one
sample was used for MRD analysis in landmark cycle 1, landmark cycle 2,
or consolidation.
Minimal Residual Disease Detection in AML
www.jco.org
© 2013 by American Society of Clinical Oncology 3891
Table 2. Univariate and Multivariate Cox Regression Analysis of Relapse Incidence for Various Prognostic Factors, Including MRD After Landmark Cycle 2 in AML
Variable
No. of
Patients
Relapse and Death in CR1 Univariate Analysis Multivariate Analysis
No. of Patients
Who Relapsed
No. of Deaths
in CR1
RFS%at
4 Years
ⴱ
Relapse %
at 4 Years
ⴱ
Death in CR1
% at 4 Years
ⴱ
HR 95% CI P HR 95% CI P
Total 183 82 11 46 48 6
MRD after cycle 2 ⬍ .001 .001
MRD ⱕ 0.1% 141 53 9 52 42 7
MRD ⬎ 0.1% 42 29 2 23 72 5 2.97 1.88 to 4.70 2.60 1.49 to 4.55
AML type .003 .054
De novo 155 63 11 50 43 7
Secondary 15 12 0 13 87 0 3.47 1.86 to 6.47 2.68 1.33 to 5.42
MDS 13 7 0 27 73 0 1.14 0.52 to 2.48 0.65 0.27 to 1.56
CR achieved ⬍ .001 .021
After cycle 1 137 52 9 50 43 7
After cycle 2 46 30 2 30 65 4 2.44 1.55 to 3.84 1.67 1.00 to 2.79
Risk of AML .001 ⬍ .001
Good 52 16 1 61 37 2
Intermediate 107 50 8 43 49 8 1.81 1.03 to 3.18 2.55 1.32 to 4.94
Poor 24 16 2 25 67 8 3.75 1.87 to 7.54 4.75 2.08 to 10.83
WBC at diagnosis (⫻ 10
9
/L) .005 .0026
ⱕ 100 157 65 10 48 45 7
⬎ 100 26 17 1 31 65 4 2.31 1.35 to 3.95 2.12 1.12 to 4.01
Age, years .13 .084
ⱕ 40 61 22 4 56 37 7
⬎ 40 122 60 7 40 54 6 1.44 0.89 to 2.35 1.56 0.93 to 2.63
G-CSF .11 .084
G-CSF ⱕ 0.1% 83 42 6 35 57 7
G-CSF ⬎ 0.1% 100 40 5 53 42 5 0.70 0.45 to 1.08 0.66 0.42 to 1.06
Last consolidation treatment ⬍ .001 ⬍ .001
None 21 19 0 0 100 0
Cycle 3 40 23 0 33 67 0 0.46 0.25 to 0.87 0.52 0.25 to 1.10
Autologous SCT 54 19 2 63 33 4 0.23 0.12 to 0.46 0.19 0.09 to 0.41
Allogeneic SCT 68 21 9 54 33 13 0.23 0.12 to 0.44 0.19 0.09 to 0.40
NOTE. In the univariate analysis, results are given for the relapse end point for each prognostic factor separately. In the multivariate analysis, all variables shown in the left-most column were included. Consolidation
treatment was included in the regression analyses with time-dependent covariates.
Abbreviations: AML, acute myeloid leukemia; CR, complete response; CR1, first CR; G-CSF, granulocyte colony-stimulating factor; HR, hazard ratio; MDS, myelodysplastic syndrome; MRD, minimal residual
disease; RFS, relapse-free survival; SCT, stem-cell transplantation.
ⴱ
Actuarial estimates at 4 years of RFS and cumulative incidence of competing risks relapse and death in CR1.
Terwijn et al
3892
© 2013 by American Society of Clinical Oncology
J
OURNAL OF CLINICAL ONCOLOGY
