From www.bloodjournal.org by guest on June 17, 2017. For personal use only.
Trisomy of Leukemic Cell Chromosomes 4 and 10 Identifies Children With
B-Progenitor Cell Acute Lymphoblastic Leukemia With a Very Low Risk of
Treatment Failure: A Pediatric Oncology Group Study
By Michael B. Harris, Jonathan J. Shuster, Andrew Carroll, A. Thomas Look, Michael J. Borowitz, William M. Crist,
Ruprecht Nitschke, Jeanette Pullen, C. Philip Steuber, and Vita J. Land
To account for the superior prognosis of hyperdiploid, B-progenitor acute lymphoblastic leukemia (ALL), we investigated
the influence of trisomy in 1021 children 2 1 year old by
recursive partitioning analysis. The patients were treated
according to a stratified, randomized study testing antimetabolite-based therapies. Trisomies of several individual chromosomes were associated with a better prognosis in a univariate statistical analysis. Of greater importance, trisomy of
both chromosomes4and 10 identified a subgroup of patients
(n = 180) with an extremely favorable 4-year event-free
survival (EFS). Combined trisomy of chromosomes 4 and 10
retained its prognostic significance after stratification of
patients by DNA index, age, and leukocyte count. Among
patients with a DNA index greater than 1.16, patients with
trisomies of both chromosomes 4 and 10 had a 4-year EFS of
96.6% (n = 161, SE = 3.8%). whereas patients with neither
or only one of these trisomies had a 4-year EFS of 70.4%
(n = 73, SE = 11.5%). All 19 patients with a DNA index 51.16
but with trisomies of chromosomes 4 and 10 remain in
remission, suggesting that favorable chromosome trisomy
dominates in a situation in which the cellular DNA content of
51.16 predicts a less favorable outcome. We conclude that
combined trisomy of chromosomes 4 and 10 independently
predicts EFS among children with B-progenitor ALL. Patients
within the B-progenitor group who have this feature (about
20% of those with clonal abnormalities)are likely to be cured
with antimetabolite-basedchemotherapy-an approach that
should produce few significant late effects.
o 1992 by TheAmerican Society of Hematology.
T
identified as independent predictors of treatment outcome
in childhood ALL.’-” Hyperdiploid ALL ( > 50 chromosomes) has been shown to confer a favorable prognosis,
while translocations such as the t(9;22) and t(1;19) have
been related to a suboptimal ~ u t c o m e . ~ When
, ~ , ~ consid~,~~
ered together with clinical risk factors, hyperdiploidy was
associated with 4 year event-free survival (EFS) of 75% to
80% among children with B-progenitor
In an effort to refine our observations on the prognostic
importance of trisomy of leukemic cell chromosomes9 in the
context of effective antimetabolite-based therapy, we examined the leukemia cell karyotypes of a large group of
children with B-progenitor cell ALL and correlated the
results with treatment outcome. We examine here, by
recursive partitioning methods, which specific trisomies
contribute independent prognostic significance, thereby
identifying a subgroup of patients who are likely to benefit
from such therapy.
Recursive partitioning and amalgamation analysis attempts to determine which variables in a set of binary
(yes/no) indicators have independent prognostic importance. The method uses the following steps. In the first
partition attempt, the patient sample is analyzed according
to a prospectively defined criterion to determine the singly
most significant prognostic factor, assuming that at least
one significant factor can be found. This will subdivide the
patient sample into two subgroups, those with versus those
without this factor. In the second partition attempt, each of
these two subgroups is then individually analyzed in exactly
the same way to attempt to subdivide each group. This
second step could lead to the creation of two groups (if no
variable is prognostically significant in either subgroup);
three groups (if one of the subgroups has prognostically
significant variables while the other does not); or four
groups (if both subgroups have prognostically significant
variables). In the first amalgamation attempt, we compare
each pair of adjacent groups, if more than two groups are
defined, starting with the group at the highest risk, and
attempt to combine groups if there is no Significant differ-
HE RECENT MARKED improvement in cure rates
for children with acute lymphoblastic leukemia (ALL)
has focused attention on better methods of identifying
patients with increased or decreased risk of treatment
failure. The aim of such analyses is to further the development of risk-directed therapy-an especially important goal
in children, whose rapid growth makes them highly susceptible to treatment-related morbidity. Unfortunately, most
prognostic factors depend on treatment
so that
no single factor or combination of factors consistently
identifies patients who can be assigned to minimal treatment with confidence in a successful outcome.
Over the past decades, certain genetic features of leukemic cells, including chromosome number (ploidy) and
nonrandom recurring chromosomal translocations, were
From the Tomorrows Children S Institute at Hackensack Medical
Center, Hackensack, NJ; the Department of Statistics of the University
of Florida and POG Statistics Office, Gainesville, FL; the Department
of Medical Genetics at the University of Alabama, Birmingham, AL;
the Department of HematoloolOncology, St Jude Children S Research Hospital and the University of Tenessee, Memphis, TN; the
Department of Pathology, Duke University Medical Center, Durham,
NC; the Department of Pediatrics, Oklahoma University, Oklahoma
City, OK; the Department of Pediatrics, University of Mississippi,
Jackson, MS; the Department of Pediatrics, Baylor College of Medicine, Houston, TX; and the Department of Pediatrics, Washington
University School of Medicine, St Louis, MO.
Submitted November 18, 1991; accepted February 24, 1992.
Supported in part by grants fi-om the National Cancer Institute and
the National Institute of Health (CA29139, CA25408, CA31566,
CA15525, CA11233, CA15989, CA03161, CA05587, and CA30969)
and the American Lebanese Syrian Associated Charities (ALSAC).
Address reprint requests to Vita J. Land, MD, Pediatric Oncology
Group, 4949 WPine Blvd, Suite 2,St Louis, MO 63108.
The publication costs of this article were defrayed in part by page
charge payment. This article must therefore be hereby marked
“advertisement” in accordance with 18 U.S.C.section 1734 solely to
indicate this fact.
0 I992 by The American Society of Hematoloo.
0006-4971192/7912-0027$3.00/0
3316
Blood, Vol 79,No 12 (June 15), 1992:pp 3316-3324
From www.bloodjournal.org by guest on June 17, 2017. For personal use only.
3317
EXTRA CHROMOSOMES 4 AND 10 IN ALL
ence in outcome. Newly defined groups are subjected to
partition attempts followed by amalgamation attempts until
no further partition or amalgamation meet the significance
criterion set forth in the algorithm. The total patient
population is subdivided into two distinct groups before
conducting the analysis. The first group (50% of the
patients) is used to build the test model as given above. The
second or verification dataset is used for inferences about
the model. It is the verification process that keeps the
studywise error rate under control, because very few statistical tests are conducted on the verification dataset. Additional technical information on this approach appears in
the Statistical Considerations.
MATERIALS AND METHODS
Patients and treatment. Between February 1986 and January
1991, 1950 children or adolescents with B-progenitor cell ALL
were entered on the Acute Leukemia in Childhood 14 (ALinC-14)
treatment protocol of the Pediatric Oncology Group (POG).
Patients 2 1year of age and 121 years with myeloperoxidase (or
Sudan black B)-negative acute leukemia, whose leukemic cells
lacked T-cell-associated surface markers and surface Ig (see
section on Immunophenotypingbelow), were eligible for this study.
Nine hundred twenty-nine patients were not assessed, for the
following reasons: too early to evaluate, 24; ineligible for study, 18;
not evaluable for response, 12; unsatisfactory or lost karyotype,
445; and normal karyotype, 430. Thus, 1,021 evaluable patients
with abnormal karyotypeswere identified in this analysis (Table 1)
and form the basis for the analyses that follow. The cutoff for
analysiswas May 2,1991.
Details of the four treatment regimens, including drug schedules
and doses, have been published previously and are summarized in
Fig l.15,16
Briefly, induction therapy (see legend to Fig 1) comprised
vincristine, prednisone, and asparaginase with triple intrathecal
chemotherapy (methotrexate, hydrocortisone, cytarabine), followed by intensificationwith: regimen A, moderate-dosemethotrexate (MTX); regimeh B, moderate-dose MTX plus asparaginase;
regimen C, moderate-dose MTX plus cytarabine administered over
16 weeks; and regimen D, moderate-dose MTX plus cytarabine
administered over 14 months. Continuation therapy comprised
daily mercaptopurine and weekly methotrexate until 3 years from
diagnosis;vincristine and prednisone pulses as systemic intensification, and triple intrathecal therapy to prevent relapse in the central
nervous system (CNS) were administered for 2 years.
Risk-group assignment was primarily based on age, leukocyte
count (Table 2 and Pullen et all4), as well as on immunophenotype
(ie, early pre-B v pre-B), because of the expected inferior prognosis
of pre-B versus early pre-B patients.’* Patients at lower risk of
relapse with early pre-B ALL were randomized to receive one of
the four regimens, whereas higher-risk patients received only
Table 1. Selection of Patientsfor Analysis
No. of
Patients
Total no. enrolled in ALinC-14 study
Total no. not assessed
NOfollow-Up
Ineligible
Not evaluable for response
Unsatisfactoryor lost karyotype
Normal karyotype
Total no. with abnormal clones (assessable)
1,950
929
24
18
12
445
430
1,021
regimens B, C, and D. Patients with pre-B ALL were randomized
only to regimens B or C because of the limited size of this cohort.
The treatment protocol was approved by the appropriate institutional review boards and the National Cancer Institute. Signed
informed consent was obtained for all patients’ participation.
Cytochemical studies. Bone marrow cells obtained at diagnosis
were stained according to standard techniques, including the use of
Wright-Giemsa, myeloperoxidase or Sudan black B, naphthol
AS-D chloroacetate esterase, and a-naphthyl butyrate esterase.
The diagnosis of ALL or acute myelogenous leukemia (AML) was
based on morphologic and cytochemical criteria of the FrenchAmerican-British (FAB) Cooperative Group.” Thus, by definition,
all cases of ALL had fewer than 3% blast cells positive for
myeloperoxidase or Sudan black B (myeloid pattern), or fewer
then 20% positive for butyrate esterase (myeloid pattern); none of
the cells had Auer rods present.
Immunologic cell fyping. Bone marrow cells were separated on
a Ficoll-Hypaque gradient. Cell surface antigens were detected by
a standard indirect immunofluorescence assay with monoclonal
antibodies to lymphoid-associated antigens. Cells were analyzed
for fluorescent activity by flow cytometry (FACScan; Becton
Dickison, San Jose, CA). Blast cells were also tested for surface Ig
(slg) and cytoplasmic Ig (clg) using a visual immunofluorescence
assay. Depending on the pattern of reactivity, cells were classified
as T (positive for at least two of the following antigens: CD7, CD5,
and CD2), B(sIg+), pre-B(cIg+), or early pre-B (CD19+, CD22=,
CD24*, CDlO’, cIg-, sIg-, CD7-, CD5-, and CD2-). Patients
were classified as having pre-B cell ALL if 2 10% of their bone
marrow blast cells contained cIg. Only patients with pre-B or early
pre-B ALL and age greater than 1year at diagnosis were enrolled
in the ALinC-14.19s20
Cytogenetic analysklDNA frow cytomehy. Bone marrow samples were processed by standard cytogenetic methods. Cases were
classified on the basis of numerical chromosomal abnormalities
(gain or loss of whole chromosomes) or structural abnormalities
(translocations, deletions, duplications, isochromosomes, inversions, and unclassifiable markers), according to the International
System for Human Cytogenetic Nomenclature.21 Cellular DNA
content determinations were made by flow cytometry to derive the
DNA index (ratio of DNA content in leukemic Go/GI cells to that
in normal diploid cells),” which was then used to distinguish to
prognostic categories of leukemic cells: those with DNA indexes
11.16 or > 1.16 (approximately equal to <53 or 2 5 3 chromosomes).
Definitions. Complete remission was defined as less than 5
mononuclear cells/pL of cerebrospinal fluid and 15% lymphoblasts in marrow aspirates; evidence of regeneration of normal
marrow cells; absence of symptoms and signs attributable to
leukemia; and a clinical performance status that had largely
returned to full function. EFS was defined as the interval between
attainment of a complete response and the earliest of the following
events: relapse, death from any cause, or date of last contact.
Patients who failed to enter remission were considered failures at
time zero.
CNS leukemia was diagnosed on the basis of cranial nerve
palsies, with or without leukemic blasts in the cerebrospinal fluid,
or when mononuclear cell counts of 2 5 cells/kL and leukemic
blasts were present in Wright-stained cytocentrifuged samples of
spinal fluid. The diagnosisof testicular leukemia required confirmation by open biopsy.
Statistical considerations. The hazard ratio is the ratio of
instantaneous risk of treatment failure for patients with that
specific trisomy to the instantaneous risk of failure in the absence
of that trisomy. EFS was the dependent variable in all analyses.
Twenty-four independent variables were defined according to
From www.bloodjournal.org by guest on June 17, 2017. For personal use only.
HARRIS ET AL
3318
Wk
4
INDUCTION ,
CNS
CONSOLIDATION
7
,
Intensification
r
RegA
156
CONTINUATION
Pulses
CNS
Triple IT x 2 on
days 29 1L 36
6 MP PO daily
(Days 29-43)
IDM
q 3 wks x 6
6 MP PO dally
MTX IM q wk
(TOwk 156)
PRED PO x 7 days
VCR days 1 + 8
9 2 mo
Same as
Reg A
Same as
Reg A
plus ASP
q wk x 24
Same as
Reg A
Same as
Reg A
Same as
Reg A
Same as
Reg A
IDMIARA-C
4 3wks x 6
Same as
Reg A
Same as
Reg A
Same as
Reg A
Same as
Reg A
IDMIARA-C
zi 12wks x 6
Same as
Reg A
from wk 70
Same as
Reg A
Same as
Reg A
PRED, VCR
and
ASP
TIT(days 0,22)
TIT
Fig 1. Schema of therapy, ALinC 14 (POG 8602) treatment protocol for patients with newly diagnosed non-1, non-&ell ALL, excluding infants
(c12 months of age). Remission induction: regimens A through D: prednisone (PRED), 40 mg/m2 day orally (maximum dose, 60 mg) for 29 days;
vincristine (VCR), 1.5 mg/m2 IV once weekly for 4 weeks; L-asparaginase(ASP), 6,000 U/m2 intramuscularly (IM)three times weekly for 2 weeks;
6-mercaptopurine (6-MP), 75 mg/ m2orally for 14 days (days 29 t o 43). Triple intrathecaltherapy (TIT) with hydrocortisone 15 mg, MTX 15 mg, and
cytosine arabinoside (AM-C) 30 mg for ages greater than 9 years (scaled down for younger children); administered on days 1and 22 of inductlon
treatment, days 29 and 36 of CNS intensification, and thereafter on weeks 9,12,15, and 18, and every 8 weeks through week 105. Intensification:
regimen A: intermediate-dose MTX (IDM), 1 g/m2 over 24 hours with leucovorin rescue weeks 7, 10,13, 16, 19, and 22; regimen B: same as
regimen A plus ASP, 25,000 U/m2 IM weekly from week 7 t o 30; regimen C: exactly as regimen A plus ARA-C 1 g/m2 over 24 hours, overlapping
with IDM by 12 hours; regimen D: IDM plus ARA-C at same doses as in regimen C but administered every 12weeks for six courses (weeks7,19,31,
43,55, and 67) with daily oral 6-MP and weekly IM MTX beginning 3 weeks after and continuing until 2 weeks before each IDM/ARA-C course.
Continuation therapy: regimens A, B, and C: MTX 20 mg/m2IM weekly and daily 6-MP orally weeks 25 t o 156; regimen D: 6-MP and MTX weeks
70 t o 156. Pulses: PRED and VCR at same doses as administered in induction; VCR weekly for 2 weeks and PRED daily orally for 7 days at weeks 8,
17,25,41,57,73,89,
and 105. (Reprinted with permission.I6)
whether trisomy of a specific chromosome was present or absent.
Recursive partitioning and amalgamation analysis, using a modification of Ciampi et al,’ similar to that of Trueworthy et al,I3 and to
that of Shuster et al,24was used. The data were divided into two
subsets: those with even POG identification numbers served as the
test (modeling) set, whereas those with odd POG identification
numbers served as the verification data set. Because recursive
partitioning and amalgamation entails a large number of statistical
tests, the verification process is vital as protection against spurious
findings.
In recursive partitioning and amalgamation analysis, a subgroup
is partitioned if an independent variable has a P value of less than
1%, based on univariate Cox regression analysis.25The variable
Table 2. Criteria for Risk Assignment
Feature
Leukocyte count (x109/L)
> 10
10-99
2 100
Liver or spleen below umbilicus; extramedullary leukemia
1-2.9
A
B
B
B
3-5.9
A
A
B
8
6-10.9
> 11
A
B
B
B
B
B
B
B
Abbreviations: A, lower-risk group; B, higher-risk group.
chosen for the partition is the one with the smallest estimated
hazard ratio among those meeting the P < .01 criterion. Amalgamation is attempted after each partitioning procedure by comparing
adjacent risk groups (starting with the highest) by univariate Cox
regression analysis, and combining them if P is > .05. The procedure is restricted to the test data set, starting with all patients in the
subset, and continuing until no further partitions or amalgamations
can be completed. The P values for each partition and amalgamation, together with comparison of the final resulting stages, are
provided for the test and verification data sets. It is within the
verification subset that inferences about the model should be
made.
This method of analysis has advantages and disadvantages
compared with the more traditional stepwise Cox regression. Most
important is the ability to detect subset specific factors. The
stepwise Cox model assumes that the log of the true hazard ratio is
linear in the covariates. This means that the hazard ratio for one
trisomy is assumed to be the same, irrespective of the value of other
covariates previously entered in the model. The stepwise Cox
model may therefore find a significant factor in a very good risk
group, only as an artifact of its significance in a poor-risk group.
While the hazard ratio estimates of both methods require proportional hazards assumptions to be valid, the P values of recursive
partition are valid at every step of the recursive partition and
amalgamation process as a test of the null hypothesis of no
prognostic importance. Only the first step of the stepwise Cox
From www.bloodjournal.org by guest on June 17, 2017. For personal use only.
3319
EXTRA CHROMOSOMES 4 A N D 10 IN ALL
procedure has this property. The recursive partitioning method
uses smaller and smaller modelling data sets as the process
continues, whereas the stepwise Cox procedure uses the entire
data set to establish each significant variable. This is an advantage
to the stepwise Cox model, if indeed a linear model in the log
hazard ratios is valid. In such a situation, the stepwise Cox model
will have a higher power of detecting truly prognostic trisomies.
Our combined use of the hazard ratio and P value, rather than
relying only on the P value as most other analysts do, is to allow
greater possibility of smaller groups (rarer trisomies) being entered
into the model. The stepwise Cox model, which relies strictly on P
values, will be much more dependent on the mix of trisomies (how
close to 50% have the trisomy?) than the method that combines
both hazard ratio and P value. This is especially important if
outstanding prognosis subgroups are sought. The hazard ratio has
the same interpretation irrespective of the mix of trisomies.
Nonetheless, both methods will have a difficult time entering very
rare but outstandingly prognostic trisomies in their respective
models.
The verification process substantially reduces the chances of
spurious findings, but also reduces the data available for modelling
by 50%. It should only be used if patient numbers permit. If all
patients are used for modelling, there is a higher probability of
detecting truly significant trisomies, as well as a higher probability
of falsely detecting truly nonprognostic trisomies as prognostic.
Stepwise procedures are good exploratory tools for finding a set
of independent prognostic indicators. Generally, after adjusting for
the variables in the model, no other variables will be significantly
prognostic. However, there may be other sets of explanatory
variables that are as predictive or even more predictive than those
derived in the stepwise process. In other words, no uniqueness can
be assumed concerning the results of a stepwise procedure. For
example, a small difference in P values could make a major
difference as to which variable enters first into the stepwise Cox
model. The final model resulting from these alternate first steps
could be very different, although both might explain differences in
outcome very well.
EFS curves were constructed by the method of Kaplan and
Meier,26with standard errors as described by Pet0 et al,27 for the
final risk groupings of the test and verification data sets, respectively. Stratified logrank tests were used to analyze the independent prognostic importance of (1) trisomy and risk groups defined
by Trueworthy et all3 and (2) trisomy and treatment.
RESULTS
The first step of the recursive partitioning analysis is
shown in Tables 3 and 4. With use of the test data set, it can
be seen that trisomies of chromosomes 4,5,6,10,14,17,18,
21, and X are univariately associated with a superior EFS
(P < .01). Of these abnormalities, trisomy 10 had the
smallest estimated hazard ratio (0.16) and, therefore, was
selected as the primary partition. The full details of
recursive partitioning appear in Tables 5 and 6. Among
trisomy 10 patients, trisomy 4 was the strongest prognostic
variable, whereas within the group of patients lacking
trisomy 10, no specific trisomy met the P < .01 criterion.
Two patient subgroups (1) without trisomy 10 or (2) with
trisomy 10 but without trisomy 4 could be amalgamated, but
no further partition based on trisomy could be made.
The recursive partition resulted in two subgroups: one
with trisomy of both chromosomes 4 and 10, and one with
trisomy of, at most, one of these two chromosomes. This
Table 3. First Step of Recursive Partition Analysis
Test Data
Without Trisomy
Verification Data
With Trisomy
Trisomy
N
Fail
Expected
N
Fail
Expected
+1
+2
+3
+4
+5
+6
+7
i 8
+9
+10
+11
+12
+13
+14
+15
+16
i17
+18
+19
+20
+21
+22
474
474
468
317
429
304
467
412
450
339
459
455
464
306
468
464
334
325
475
474
231
454
275
454
106
105
106
93
103
84
104
96
103
99
105
104
105
86
106
105
90
95
103
104
72
100
76
101
104.4
101.2
102.6
68.8
94.7
65.3
102.8
91.9
98.6
73.5
99.3
99.2
101.1
68.4
101.8
100.7
72.6
70.4
104.2
104.0
48.0
99.1
60.9
100.6
8
18
14
165
53
178
15
70
32
143
23
27
18
176
14
18
148
157
7
8
251
28
207
28
0
1
0
13
3
22
2
10
3
7
1
2
1
20
0
1
16
11
3
2
34
6
30
5
1.6
4.8
3.4
37.2
11.3
40.7
3.2
14.1
7.4
32.5
6.7
6.8
4.9
37.6
4.2
5.3
33.4
35.6
1.8
2.0
58.0
6.9
45.1
5.4
+X
+Y
Without Trisomy
HR
95%Conf(HR)
0.26
0.24
0.42
0.14-0.46
0.08-0.77
0.26-0.67
0.68
0.35-1.30
0.16
0.07-0.34
0.42
0.26-0.69
0.39
0.23
0.23-0.66
0.12-0.43
0.39
0.26-0.59
0.53
0.35-0.81
With Trisomy
N
Fail
Expected
N
Fail
Expected
533
525
527
375
493
374
523
459
502
381
506
516
523
354
523
519
386
371
534
527
267
513
323
520
100
101
100
94
95
86
101
92
99
94
100
101
98
90
98
100
92
94
100
99
76
100
81
98
99.7
97.5
97.5
69.6
90.8
70.0
97.6
85.2
93.4
68.8
94.4
96.5
95.6
66.3
97.5
95.7
71.6
68.7
99.8
97.4
47.9
94.0
60.6
98.1
6
14
12
164
46
165
16
80
37
158
33
23
16
185
16
20
153
168
5
12
272
26
216
19
1
0
1
7
6
15
0
9
2
7
1
0
3
11
3
1
9
7
1
2
25
1
20
3
1.3
3.5
3.5
31.4
10.2
31.O
3.4
15.8
7.5
32.2
6.6
4.5
5.4
34.7
3.5
5.3
29.4
32.3
1.2
3.6
53.1
7.0
40.4
2.9
HR
95%Conf(HR)
0.16
0.56
0.39
0.08-0.36
0.24-1.28
0.23-0.68
0.53
0.26-1.04
0.16
0.07-0.33
0.23
0.12-0.44
0.24
0.16
0.12-0.47
0.07-0.34
0.30
0.19-0.46
0.37
0.23-0.60
Abbreviation: HR, estimated hazard ratio (ratio of risks o f treatment failure for patients with, to that of those without, the specific trisomy).
From www.bloodjournal.org by guest on June 17, 2017. For personal use only.
3320
HARRIS ET AL
Table 4. Univariate Two-sided P Values for EFS
Trisomy Data
Test Data
.20
,076
,062
< ,001
,009
< ,001
.48
.24
,092
< .001
,023
,058
,072
< ,001
,037
,056
<.001
< ,001
.40
.98
< ,001
.72
.003
.84
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
X
Y
subgrouping was prognostically significant in the verification data set (P < .001). In fact, simultaneous trisomy of
chromosomes 4 and 10 was prognostically significant in the
verification data set (P < .05) for each treatment regimen
except A,to which patient entry was restricted to the group
at lower risk of treatment failure. The overall P value in the
verification set, stratified for treatment, was < .001, indicating that the prognostic significance of these trisomy groups
cannot be attributed to therapeutic differences.
Trueworthy et all3 established a three-level risk grouping
for B-progenitor cell ALL, as follows: (1) DNA index
greater than 1.16; ( 2 ) DNA index I1.16, age I 11.0 years,
and leukocyte count less than 50 x lo9& (3) DNA index
< 1.16, age greater than 11.0 years, and/or leukocyte count
2 50 x 109/L.After stratifying for the above risk grouping,
none of the variables studied by Trueworthy et all3 was
prognostically significant. Table 7 presents the analysis of
the prognostic importance of trisomy 4 and 10 within these
risk groups. Table 8 considers the independent prognostic
importance of a risk grouping based on age, leucocyte
count, and DNA index,13 after adjustment for trisomy of
chromosomes 4 and 10. Simultaneous trisomy of 4 and 10
was strongly prognostic, even within patient groups with a
DNA index greater than 1.16, the subgroup established by
Trueworthy et all3 as having an outstanding prognosis. As
shown in Fig 2, a DNA index greater than 1.16 without
Verification Data
.79
,056
.I8
< ,001
.17
< .001
.061
.061
,036
< ,001
,024
,031
.29
< ,001
.79
,055
<.001
< ,001
.87
.39
< ,001
.19
< .001
.95
Table 5. Recursive Partition Test Data Set Used to Build Model
Step 1: Partition
Patient Subgroup
Without Trisomy 10
N
Fail
Patient Subgroup
With Trisomy 10
Expected
N
Fail
Expected
HR
73.5
68.8
143
158
7
7
32.5
32.2
0.16
0.16
Test
339
99
Verification
381
94
P < .001(test and verification data sets)
95%
Confidence
0.07-0.34
0.07-0.33
Step 2: Partition Within Patient Subgroup With Trisomy 10: Patient Subgroup With Trisomy4
Patient Subgroup
Without Trisomy 4
N
Fail
Patient Subgroup
With Trisomy 4
Expected
Test
27
5
1.1
36
5
1.9
Verification
P < .001(test data set), P = ,009(verification data set).
The patient subset without trisomy 10 could not be partitioned.
N
Fail
Expected
HR
116
122
2
2
5.9
5.1
0.075
0.15
95%
Confidence
0.015-0.39
0.029-0.78
Step 3: Amalgamation Attempt
Patient Subgroup
With Trisomy 10
Without Trisomy 4
Fail
N
Test
27
5
Verification
36
5
P = .36(test data set), P = .046(verification data set).
This amalgamation attempt suceeded (test data used to model).
Patient Subgroup
Without Trisomy 10
Expected
N
Fail
Expected
7.4
11.3
339
381
99
94
96.6
87.7
Step 4: Partition Attempt
~
No trisomy met criterion.
Final grouping (P< ,001in verification data set A v 6).
Group 1:Trisomyfor both chromosomes 4 and 10.
Group 2:Trisomy of at most one of chromosomes 4 and 10.
From www.bloodjournal.org by guest on June 17, 2017. For personal use only.
3321
EXTRA CHROMOSOMES 4 AND 10 IN ALL
Table 6. Results of Final Stages Defined by Recursive PartitioningAnalysis: Test and Verification Data Sets
Test Data
Verification Data
Group
Definition
4-yr EFS
SE
N
4-yr EFS
SE
N
1
2
Trisomy of chromosomes 4 and 10
Trisomy of only one or none of chromosomes 4 and 10
97.5%
60.3%
4.1 %
6.6%
116
366
95.3%
58.2%
4.9%
6.7%
122
417
Comparison of EFS for group 1 v group 2 (test data): P < ,001. Comparison of EFS for group 1 v group 2 (verification data): P < .001.
trisomy of both chromosomes 4 and 10 is not associated
with a superior outcome (4-year EFS of 70.4% 2 11.5% v
an overall 4-year EFS of 72.4% 2 2.6% among all 1,896
eligible patients).
To describe the patient subgroups, clinical information is
as follows. Of the 783 patients with at most one trisomy of
chromosomes 4 and 10, 50% had translocations, median
chromosome count was 46, median leukocyte count was
11.9 x 109/L, and median age was 4 years. Of the 238
patients with trisomy of both chromosomes 4 and 10, 18%
had translocations, median chromosome count was 56,
median leukocyte count was 6.0 x 109/L, and median age
was 4 years. By these clinical features, one would expect the
group with trisomy of both 4 and 10 to have a better
prognosis than those with at most one of these trisomies.
However, these clinical features alone cannot explain the
extent of the difference in EFS.
DISCUSSION
The most important finding of this study is that information obtained from the leukemia cell karyotype permits
accurate identification of a subgroup of children or adolescents with B-progenitor cell ALL who appear curable with
antimetabolite-based chemotherapy. Prospective identification of this low-risk group should stimulate the development of risk-specific therapy in future studies, thus leading
to an improved quality of life both during and after therapy,
while preserving a high rate of cure.
That patients with trisomy of chromosomes 4 and 10 have
a 4-year EFS of approximately 95% is remarkable. However, a current limitation to the use of trisomy as a
prognostic factor is that the karyotype cannot be consis-
tently determined in all newly diagnosed cases of ALL. The
proportion of patients having an identifiable abnormal
clone was just over one-half in this study (1,021 of 1,896,
54%), which relied on a central reference laboratory for
karyotype determination. This result contrasts with the
90% detection rate reported by Williams et ala from a
single institution. A selection bias may thus be present in
our study, and results can be considered valid only for
patients with identified clonal abnormalities. For example,
patients with a clonal abnormality had a 4-year EFS of
66.9% (n = 1,021; SE = 3.9%) as compared with the remaining eligible patients’ 4-year EFS of 78.8% (n = 875;
SE = 3.2%; P = .018).
The biologic explanation for these prognostic associations is not entirely clear. The patient group with combined
trisomy of chromosomes 4 and 10 appear to have more
prognostically favorable clinical features. Even so, the
favorable correlation with EFS of trisomy of chromosomes
4 and 10 cannot be attributed entirely to a higher frequency
of lower-risk clinical features, because the presence of such
trisomy retained prognostic importance within the groups
defined by DNA index, age, and white blood cell count
(WBC), the complete set of independently prognostic
indicators found by Trueworthy et all3 (Table 7).
The strong association of hyperdiploidy of leukemic cell
chromosomes with favorable treatment outcome was originally reported by Secker-Walker et all0 and since confirmed
by other^.^,^,^^ Patients whose leukemic cell karyotypes
contain greater than 51 or 52 chromosomes have been
reported to have a more favorable prognosis than those
with lower ploidy,’ although a precise causal explanation
for this finding is lacking. Further, the DNA index deter-
Table 7. Prognostic Importanceof Trisomy Defined Risk Groups
Patient Subgroup With Trisomy
of Neither or One of Chromosomes
4 and 10
Risk group
1
2
3
DNA index not
available
Risk group 1: DNA index
2: DNAindex
3: DNAindex
Patient Subgroup With Trisomy
of Chromosomes4 and 10
N
Fail
Expected
N
Fail
Expected
PValue
73
455
149
106
11
105
65
22
4.5
99.9
64.4
16.1
161
17
2
58
2
0
0
2
8.5
5.1
0.6
7.9
< .001
.02
,010
> 1.16
<1.16(age s11.0yrand WBC <50 x 10g/L)
s1.16(age >ll.Oyrand/orWBC >50 x 109/L)
The chromosome prognostication is an independent factor over and above ploidy. Pvalue is .0127 for prognostic importance of trisomy of 4 and 10
in groups 2 and 3 combined (DNA index 5 1.16).
From www.bloodjournal.org by guest on June 17, 2017. For personal use only.
HARRIS ET AL
3322
Table 8. Prognostic Importance of Leucocyte Count, Age, and DNA Index Defined Risk Groups
Group 1
Trisomy of:
At most one of the chromosomes 4 and 10
Both chromosomes 4 and 10
Group 3
Group 2
N
Fail
Expected
N
Fail
Expected
N
Fail
Expected
73
11
22.0
455
105
126.2
1 49
65
32.9
161
2
1 .?
17
0
0.3
2
0
0.0
~~
P < ,001within group with at most one of trisomy 4 and 10.P = not applicable within group with trisomy of both chromosomes 4 and 10.Risk
group 1, DNA index >1.16;2, DNA index 51.16 (age 511.0yr and WBC <50 x lOg/L); 3,DNA index 51.16 (age >11.0yr and/or WBC >50 x
10~~).
mined by flow cytometry (ratio of leukemia cell DNA
content of Go/GI cells to that of normal diploid cells),
which can be done cheaply, rapidly, and with near uniform
success, provides a good estimate of the chromosome
number per cell ( 2 2 whole chromosomes) and has also
been strongly associated with prognosis. The rate of successful DNA flow cytometry in the present study (84% of cases
with a detectable clonal abnormality) is lower than might be
expected because a DNA index was determined only if
marrow remained after the mandatory immunophenotyping and cytogenetic studies. When a sample was submitted
for DNA analysis, the success rate was 99%. Use of DNA
ploidy alone has enabled investigators to identify a subgroup of patients with a 4-year EFS of 75% to 9O%.l2J3
We previously reported that hyperdiploidy greater than
51 was associated with an improved treatment outcome and
that trisomy of several autosomes, including 6, 12, and 21,
was associated with a better EFS within the greater than 51
chromosome group. We were unable to identify a cohort of
patients whose prognosis was as favorable as that for the
trisomy 4 and 10 subgroup in the present study.9This failure
can be explained partly by the lower success rate and the
poorer quality of cytogenetic studies in the first investigation, in which a central reference laboratory was not
available, and by the generally less effective treatment
regimens that were used and were associated with significantly higher rates of bone marrow relapse. Moreover,
there were no efforts made to obtain DNA indices. In a
.... ...........
80
....___.._......
.....
L..
- - - Rlscmy 4 a 10 and DNA index Sl.W
------- " m y 4 h and DNA index >l.*
(A)
(8)
x)
8
P
0
1
2
3
5
4
8
Years followed
Patients at Risk
19
181
73
608
9
83
34
221
5
(A)
22
(8)
(C)
(D)
11
48
Fig 2. Probabilityof EFS for patients with (A) an unfavorable DNA
index (51.16) and trisomy of both chromosomes 4 and 10; (B) a
favorable DNA index (>1.16) and trisomy of these chromosomes; (C)
a favorable DNA index without trisomy of both of these chromosomes; and (D) an unfavorable DNA index without such trisomy.
more recent analysis, we showed the value of DNA index,
leukocyte count, and age as predictors of treatment outcome in patients with B-progenitor ALL treated on the
POG ALinC 14 study reported here.13 We have extended
our previous observations by showing the independent
prognostic importance of trisomy of chromosomes 4 and 10
among hyperdiploid cases in general, providing further
insight as to which trisomic chromosomes are most strongly
associated with the favorable prognosis consistently reported for hyperdiploid ALL. These results suggest that
increased copy number of genes located on chromosomes 4
and 10 may render leukemic lymphoblasts more sensitive to
one or more of the chemotherapeutic drugs used in this
trial, perhaps by altering intracellular drug metabolism, as
had been suggested by Whitehead et
Surprisingly,
patients with a favorable DNA index, but lacking trisomy of
chromosomes 4 and 10, fared suboptimally on POG regimens (4-year EFS of 70%), in contrast to patients with an
unfavorable DNA index but with trisomy of chromosomes 4
and 10 (all 19 remain in complete remission). This information suggests that, provided clonal abnormalities are identified, trisomy of these two autosomes should be given more
prognostic weight than a favorable DNA index when one
attempts to assign children to protocols of minimal therapy.
Optimal therapy for "lower risk" children with ALL defined in this way remains to be determined by prospective
randomized trials building upon these findings.
Our recursive partitioning method with verification selects patient subgroups on the basis of the hazard ratio for
treatment failure, not strictly because of a small P value.21
This makes it more sensitive to finding subgroups with an
excellent prognosis. Methods purely based on P values tend
to be much more dependent on the relative sizes of the
subgroups.
The results described here would be more generally
applicable if cellular genetic defects could be identified in
cases without fully banded karyotypes. Recently, fluorescence in situ hybridization (FISH) analysis has been introduced and is being widely applied in the study of genetic
abnormalities in malignancies. It can detect specificchromosomes directly on metaphase or interphase spreads of
human chromosomes using chromosome-specific probes,
and has been used to detect trisomies, particularly trisomy 9
in acute myeloblastic le~kemia.3~
Further application of
this technique should make it possible to test the predictions of a recursive partitioning model on a large group of
children with ALL.
From www.bloodjournal.org by guest on June 17, 2017. For personal use only.
3323
EXTRA CHROMOSOMES 4 AND 10 IN ALL
REFERENCES
1. Miller D R Childhood acute lymphoblastic leukemia: 1.
Biological features and their use in predicting outcome of treatment. Am J Pediatr Hematol Oncol40163,1988
2. Rivera GK, Mauer AM: Controversies in the management of
childhood acute lymphoblastic leukemia: Treatment intensification, CNS leukemia and prognostic factors. Semin Hematol 24:12,
1987
3. Crist W, Pullen J, Boyett J, Falletta J, van Eys J, Borowitz M,
Jackson J, Dowell B, Frankel I, Quddus F, Ragab A, Vietti T
Clinical and biological features predict a poor prognosis in acute
lymphoid leukemias in infants: A Pediatric Oncology Group Study.
Blood 67:135,1986
4. Crist W, Pullen J, Boyett J, Falletta J, van Eys J, Borowitz M,
Jackson J, Dowell B, Russell C, Quddus F, Ragab A, Vietti T:
Acute lymphoid leukemia in adolescents: Clinical and biological
features predict a poor prognosis: A Pediatric Oncology Group
Study. J Clin Oncol6:34,1988
5. Pui C-H, Dodge RK, Dahl GV, Rivera G, Look AT, Kalwinsky D, Bowman WP, Ochs J, Abromowitch M, Mirro J, Murphy SB:
Serum lactic dehydrogenase level has prognostic value in childhood
acute lymphoblastic leukemia. Blood 66:778, 1985
6. Reaman G, Zeltzer P, Bleyer WA, Amendola B, Level C,
Sather H, Hammond D: Acute lymphoblastic leukemia in infants
less than one year of age: A cumulative experience of the
Children’s Cancer Study Group. J Clin Oncol3:1513,1985
7. Williams DL, Tsiatis A, Brodeur GM, Look AT, Melvin SL,
Bowman WP, Kalwinsky DK, Rivera G, Dahl G V Prognostic
importance of chromosome number in 136 untreated children with
acute lymphoblastic leukemia. Blood 60864,1982
8. Bloomfield CD, Goldman AI, Alimena G, Berger R, Borgstrom GH, Brandt L, Catovsky D, de la Chapelle A, Dewald GW,
Garson OM, Garwicz S, Golomb HM, Hossfeld DK, Lawler SD,
Mitelman G, Nilsson P, Pierre RV, Philip P, Prigogina E, Rowley
JD, Sakurai M, Sandberg AA, Secker-Walker LM, Tricot G, Van
Den Berghe H, Orshoven AV, Vuopio P, Whang-Peng J: Chromosomal abnormalities identify high-risk and low-risk patients with
acute lymphoblastic leukemia. Blood 67:415,1986
9. Jackson JF, Boyett J, Pullen J, Brock B, Patterson R, Land V,
Borowitz M, Crist W: Favorable prognosis associated with hyperdiploidy in children with acute lymphoblastic leukemia correlates with
extra chromosome 6: A Pediatric Oncology Group Study. Cancer
66:1183,1990
10. Secker-Walker LM, Lawler SD, Hardisty RM: Prognostic
implications of chromosomal findings in acute lymphoblasticleukemia at diagnosis. Br Med J 21529,1978
11. Swansbury GJ, Secker-Walker LM, Lawler SD, Hardisty
RM, Sallan SE, Garson OM, Sakurai M: Chromosomal findings in
acute lymphoblastic leukaemia of childhood: An independent
prognostic factor. Lancet 2249,1982
12. Look AT, Roberson PK, Williams DL, Rivera G, Bowman
WP, Pui C-H, Ochs J, Abramowitch M, Kalwinsky D, Dahl GV,
George S, Murphy SB: Prognostic importance of blast cell DNA
content in childhood acute lymphoblastic leukemia. Blood 65:1079,
1985
13. Trueworthy R, Shuster J, Look AT, Crist W, Borowitz M,
Carroll A, Frankel L, Harris M, Wagner H, Haggard M, Mosijczuk
A, Pullen J, Steuber P, Land V Ploidy of lymphoblasts is the
strongest predicator of treatment outcome in B-progenitor cell
acute lymphoblastic leukemia in childhood: A Pediatric Oncology
Group Study. J Clin Oncol10:606,1992
14. Pullen DJ, Boyett JM, Crist WM, Falletta JM, Roper M,
Dowell B, van Eys J, Jackson JF, Humphrey GB, Metzgar PS,
Cooper MD: Pediatric Oncology Group utilization of immunologic
markers in the designation of acute lymphocytic leukemia subgroups: Influence on treatment response. Ann NY Acad Sci 428:26,
1984
15. Crist W, Carroll A, Shuster J, Jackson J, Borowitz M, Behm
F, Link M, Steuber P, Ragab A, Land V, Pullen J: Philadelphia
chromosome positive childhood acute lymphoblastic leukemia:
Clinical features and outcome. A Pediatric Oncology Group
(POG) Study. Blood 76:489,1990
16. Crist WM, Carroll AJ, Shuster JJ, Behm FG, Whitehead M,
Vietti TJ, Look AT, Mahoney D, Ragab A, Pullen DJ, Land VJ:
Poor prognosis of children with pre-B acute lymphoblastic leukemia is associated with the t(1;19)(q23;p34): A Pediatric Oncology
Group Study. Blood 76:117,1990
17. Bennett JM, Catoysky D, Daniel MT, Flandrin G, Galton G,
Gralnick HR, Sultan C: Proposals for the classification of the acute
leukemias. Br J Hematol33:451,1976
18. Crist W, Boyett J, Roper M, Pullen J, Metzgar R, van Eys J,
Ragab A, Starling K, Vietti TJ, Cooper M: Pre-B cell leukemia
responds poorly to treatment: A Pediatric Oncology Group Study.
Blood 63:407,1984
19. Gathings WE, Lawton AR, Cooper MD: Immunofluorescent studies of the development of pre-B, B-lymphocytes and
immunoglobulin isotype diversity in humans. Eur J Immunol7:804,
1977
20. Pullen DJ, Falletta JM, Crist WM, Vogler LB, Dowell B,
Humphrey GB, Blackstock R, van Eys J, Cooper MD, Metzgar RS,
Meydrech E F Southwest Oncology Group experience with immunologic phenotyping in acute lymphocytic leukemia of childhood.
Cancer Res 41:4801,1981
21. ISCN (1985): An Intemational System of Human Cytogenetic Nomenclature. Published in collaboration with Cytogenet
Cell Genet Basel, Switzerland, Karger, 1985; also in Birth Defects:
Original Article Series, vol 21, no 1. New York, NY, March of
Dimes Birth Defects Foundation, 1985
22. Look AT, Melvin SL, Williams DL, Brodeur GM, Dahl GV,
Kalwinsky DK, Murphy SB, Maurer AM: Aueuploidy and percentage of S-phase cells determined by flow cytometry correlate with
cell phenotype in childhood acute leukemia. Blood 60959,1982
23. Ciampi A, Hogg SA, Kates L Stratification by stepwise
regression, corresponding analysis, and recursive partitioning: A
comparison of three models of survival analysis with covariates.
Comput Stat Data Anal 4:185, 1986
24. Shuster JJ, Falletta JM, Pullen DJ, Crist WM, Humphrey
GB, Dowell BL, Wharam MD, Borowitz M: Prognostic factors in
childhood T-cell acute lymphoblastic leukemia: A Pediatric Oncology Group Study. Blood 79166,1990
25. Cox DR: Regression models and life tables. J R Stat SOCB
34:187,1987
26. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457, 1958
27. Pet0 R, Pike MC, Armitage P, Breslow NE, Cox DR,
Howard SV, Mantel N, McPherson K, Pet0 J, Smith PG: Design
and analysis of randomized clinical trials requiring prolonged
observation of each patient. 11. Analysis and examples. Br J Cancer
35:1,1977
28. Williams DL, Harber J, Murphy SB, Look AT, Kalwinsky
DK, Rivera G, Melvin SL, Stass S, Dahl G V Chromosomal
translocations play a unique role in influencing prognosis in
childhood acute lymphoblastic leukemia. Blood 68:205,1986
29. Bloomfield CD, Secker-Walker LM, Goldman AL, Van Den
From www.bloodjournal.org by guest on June 17, 2017. For personal use only.
3324
Berghe H, de la Chapelle A, Ruutu T, Alimena G, Garson OM,
Golomb HM, Rowley JO, Kaneko Y, Whang-Peng J, Prigogina E,
Philip P, Sandberg AA,Lawler SD, Mitelman F Six year follow-up
of the clinical significance of karyotype in acute lymphoblastic
leukemia. Cancer Genet Cytogenet 40:171,1989
30. Whitehead VM, Rosenblatt DS, Vuchich M, Shuster JJ,
Witte A, Beaulieu D: Accumulation of methotrexate and methotrex-
HARRIS ET AL
ate polyglutamates in lymphoblasts at diagnosis of childhood acute
lymphoblastic leukemia: A pilot prognostic factor analysis. Blood
7644,1990
31. Anastasi J, LeBeau MM, Vardiman JW, Westbrook C A
Detection of numerical chromosomal abnormalities in neoplastic
hemotopoietic cells by in situ hybridization with a chromosomespecific probe. Am J Pathol136:131,1990
From www.bloodjournal.org by guest on June 17, 2017. For personal use only.
1992 79: 3316-3324
Trisomy of leukemic cell chromosomes 4 and 10 identifies children
with B-progenitor cell acute lymphoblastic leukemia with a very low
risk of treatment failure: a Pediatric Oncology Group study
MB Harris, JJ Shuster, A Carroll, AT Look, MJ Borowitz, WM Crist, R Nitschke, J Pullen, CP
Steuber and VJ Land
Updated information and services can be found at:
http://www.bloodjournal.org/content/79/12/3316.full.html
Articles on similar topics can be found in the following Blood collections
Information about reproducing this article in parts or in its entirety may be found online at:
http://www.bloodjournal.org/site/misc/rights.xhtml#repub_requests
Information about ordering reprints may be found online at:
http://www.bloodjournal.org/site/misc/rights.xhtml#reprints
Information about subscriptions and ASH membership may be found online at:
http://www.bloodjournal.org/site/subscriptions/index.xhtml
Blood (print ISSN 0006-4971, online ISSN 1528-0020), is published weekly by the American
Society of Hematology, 2021 L St, NW, Suite 900, Washington DC 20036.
Copyright 2011 by The American Society of Hematology; all rights reserved.