|

Received: 27 July 2017    Accepted: 13 October 2017

DOI: 10.1111/ijlh.12760

ORIGINAL ARTICLE

Comparison of conventional cytomorphology, flow cytometry

immunophenotyping, and automated cell counting of CSF for
detection of CNS involvement in acute lymphoblastic leukemia

J. C. Jaime-Pérez  | M. F. Borrego-López | R. A. Jiménez-Castillo | 

N. Méndez-Ramírez | R. Salazar-Riojas | L. T. Fernández | D. Gómez-Almaguer

Department of Hematology, Dr. José Eleuterio

González University Hospital, School of
Medicine of the Universidad Autónoma de
Nuevo León, Monterrey, México
Correspondence
Dr. José Carlos Jaime-Pérez, Servicio de
Hematología, Hospital Universitario, “Dr. José
E. González”, Monterrey, NL, México.
Email: carjaime@hotmail.com
Abstract
Background and objective: Cytospin conventional cytomorphology (CCC) is the stand-
ard method for detecting lymphoblasts in cerebrospinal fluid (CSF) of patients with
acute lymphoblastic leukemia (ALL) and for guiding treatment decisions. We evaluated
flow cytometry immunophenotyping (FCI) performance for improving detection of
central nervous system (CNS) involvement in ALL.
Methods: This prospective study included analysis of consecutive CSF samples of pa-
tients of all ages with ALL at 3 clinical stages: new diagnosis, relapse suspicion, and
after relapse treatment. Manual, cytospin, automated, and FCI methods were com-
pared and their performance statistically assessed. Using FCI as the reference method,
optimal CSF cutoff cell count that better correlated with presence of lymphoblasts
was established by receiver operating characteristic (ROC) curve analysis.
Results: Seventy-­seven CSF samples were investigated, 35 (45.4%) from newly diag-
nosed cases, 30 (39%) suspicion of relapse, and 12 (15.6%) after treatment for relapse.
Median manual WBC count in patients with CNS involvement detected by FCI was
3.75 cells/μL (0.0-­1280), and this was also the count that best correlated with CNS
infiltration (sensitivity, 50.0%; specificity, 82.2%). Compared with FCI, CCC sensitivity
and specificity were 28.6% and 100%. Automated CSF WBC count in patients with
CNS involvement detected by FCI was 5 (0.0-­1578). For automated count, optimal
WBC cutoff was 4.5 cells/μL (sensitivity, 62.5%; specificity, 70.5%).
Conclusion: Flow cytometry immunophenotyping complements conventional cyto-
spin analysis for detection of lymphoblasts in the CSF of ALL patients at any clinical
stage.

KEYWORDS
acute lymphoblastic leukemia, central nervous system, cerebrospinal fluid, cytospin, flow
cytometry immunophenotyping

1 |  INTRODUCTION a poorer prognosis and requiring CNS-­directed therapy.1-4 Routinely,

manual cell count (MCC) and microscopic examination are per-
Central nervous system (CNS) involvement is present in 5%-­10% of formed on cerebrospinal fluid (CSF) samples by standard optical
acute lymphoblastic leukemia (ALL) patients at diagnosis, conferring microscopy in a Fuchs-­Rosenthal counting chamber. This manual

Int J Lab Hem. 2017;1–6. © 2017 John Wiley & Sons Ltd |  1
wileyonlinelibrary.com/journal/ijlh  
 |
2       JAIME-­PÉREZ et al.
procedure is time-­consuming and requires experienced technolo-
gists; these drawbacks could be improved with contemporary auto-
mated cell counters which incorporate a body fluid channel capable
of analyzing and counting cells in CSF;5 importantly, this method
does not discriminate between benign and malign cells and is not
routinely used as a diagnostic option for CNS involvement in ALL.
Cytospin conventional cytomorphology (CCC) evaluation after cy-
tocentrifugation of the CSF is the gold standard for detection of
lymphoblasts in ALL patients;6,7 this method is estimated to have
a specificity greater than 95% for CNS infiltration but a low sensi-
tivity of about 50% for identification of CNS disease.8 Recent data
demonstrate that most of CNS relapses occur in patients without
CNS involvement detected at diagnosis by CCC.9 Several studies
have demonstrated that flow cytometry immunophenotyping (FCI)
is a valuable tool that reliably detects phenotypically abnormal cells
and is more sensitive than CCC analysis of CSF;10 it has been re-
cently reported that up to 30% of patients had CNS involvement at
ALL diagnosis detected using this method.3 Although FCI is not the
gold standard method for detection of CNS leukemia, it has import-
ant advantages over cytospin morphology analysis, including the
sensitivity to identify lymphoblasts in CSF samples with very low
white blood cells (WBC) counts.
The goal of this study was to compare the performance of the dif-
ferent diagnostic methods employed in the hematology laboratory to
investigate the presence of lymphoblasts in the CSF of ALL patients at
different clinical scenarios.
2 |  PATIENTS AND METHODS
This prospective study included 77 consecutive ALL patients of all
age groups and clinical stages from August 2015 to December 2016
treated at the Hematology Department of the Dr. José E. González
University Hospital of the School of Medicine of the Universidad
Autónoma de Nuevo León in Monterrey, México. This is an academic
reference center for low-­income open population patients in the
north-­east region of the country. The study protocol was approved by
the Institutional Research Board and Ethics Committee. Cerebrospinal
fluid samples, ≥2.0 mL, were collected after diagnostic or therapeutic
lumbar puncture and CSF studied by MCC with a Fuchs-­Rosenthal
Counting Chamber (Hausser Scientific, Horsham, PA, USA); auto-
mated cell count (ACC), CCC, and FCI in 3 clinical circumstances: at
diagnosis of ALL, in ALL patients with clinical symptoms suggestive
of CNS relapse, and after completing scheduled treatment for CNS
relapse in the presence of a control CC negative sample.
2.1 | Manual cell count
Cerebrospinal fluid specimens collected from ALL patients meeting
inclusion criteria were processed immediately on receipt, adding the
sample by capillarity on the Fuchs-­Rosenthal chamber. White and red
blood cells in the CSF sample were counted with a 40 × Axiostar Plus
optical microscope (Carl Zeiss, Jena, Germany).
2.2 | Automated cell count
Cerebrospinal fluid cell counting was carried out employing the body
fluid mode of a Sysmex XT-­4000i (Sysmex Corporation, Kobe, Japan)
automated hematology analyzer, capable of analyzing CSF using
flow cytometry. Samples are gently mixed to prevent cell lysis, the
sample tube is positioned below the aspiration needle, and 85 μL are
aspirated and analyzed by flow cytometry employing the body fluid
mode, with the extended count setting to improve precision in low
cell count specimens. The body fluid analysis mode uses 4-­DIFF and
red blood cells (RBC) distribution obtained after sequential analysis
to calculate WBC, mononuclear cells (MNC), and polymorphonuclear
cells (PMN), in addition to percentages and RBC counts found in body
fluids (RBC-­BF).5
2.3 | Cytocentrifuged sample preparation
and analysis
The analysis was done in all samples at diagnosis and at each spinal
tap for intrathecal (IT) therapy. For the study, 200 μL of sample was
centrifuged onto glass microscope slides using the Wescor Cytopro
Rotor AC-­060 7620 model (Wescor, Logan, UT, USA) cytocentri-
fuge machine. Centrifuge settings for the machine were 110 g for
4 minutes. The slides were air-­dried and then stained with Wright
colorant. After staining, slides were analyzed by experimented ex-
aminers with an Axiostar Plus optical microscope at 40 ×  and 100 × 
to observe the cellular characteristics and identify cells present.
Based on this observation, they were classified into benign or ma-
lignant cells. Specimens were interpreted without knowledge of FCI
findings.
2.4 | Flow cytometry immunophenotyping
Specimens of CSF were processed immediately on receipt in the
laboratory. Cells from the samples were stained using a panel of
fluorochrome-­conjugated antibodies directed to B and T cells. Eight-­
color flow cytometry was performed on a FACSCanto II flow cytom-
eter (BD Biosciences, San José, CA, USA). Fluorochrome-­conjugated
antibodies were obtained from Becton Dickinson (BD Biosciences).
Samples of CSF (≥2.0 mL) were centrifuged (300 g × 5 minutes)
and the supernatant discarded, 2.0 mL of 0.5% albumin phosphate-­
buffered saline (PBS) was added, samples were then centrifuged
(300 g × 5 minutes), the supernatant discarded, and the cells remain-
ing suspended in 300 μL of 0.5% albumin PBS for analysis.
For B-­cell ALL, antibodies used were CD3/CD10/CD14/CD19/
CD20/CD34/CD38/CD45/CD56, and for T-­cell ALL, the combi-
nation of antibodies used was CD3c/CD3s/CD4/CD5/CD7/CD8/
CD14/CD45/CD56. The fluorochromes were FITC/PE/PerCP-­Cy5.5/
PECy7/APC/APC-­H7/Horizon V500/Horizon V450.
After staining, 2.0 mL of FACS lysing solution (BD Biosciences) was
added, and after 10 minutes of incubation at room temperature, the
samples were centrifuged at 300 g × 5 minutes, supernatant was then
discarded, 2.0 mL of PBS was added, and the samples were centrifuged
JAIME-­PÉREZ et al. |
      3

as above and suspended in 300 μL of 0.5% albumin PBS immediately T A B L E   1   General and laboratory characteristics, including clinical
before data acquisition. After gentle mixing, samples were analyzed status at the time of cerebrospinal fluid (CSF) analysis of 77 patients
employing FACSDiva software version 6 (BD Biosciences), and infor- with acute lymphoblastic leukemia (ALL)

mation of all events corresponding to nucleated cells in the stained Characteristic n (%)
sample was acquired. For data analysis, the INFINICYT software ver-
Age, median (range) 17 (2-­61)
sion 1.7 (Cytognos S.L., Salamanca, Spain) was used. Identification
Sex
of B cells was based on the expression of CD45/CD19, and T cells
 Male 42 (54.5)
were identified based on CD3c. A CSF sample was considered satis-
 Female 35 (45.5)
factory for FCI analysis when ≥50 viable events were collected and
Immunophenotype
positive for lymphoblasts if there was >0.1% cells with malignant
immunophenotype. 11  B cell 72 (93.5)
 T cell 5 (6.5)
Risk at diagnosis
2.5 | Statistical analysis
 High risk 50 (64.9)
We conducted a descriptive analysis based on the clinical records and  Standard risk 27 (35.1)
regular follow-­up of each patient and built a database using SPSS ver- Clinical stage at time of CSF analysis
sion 22.0 (IBM SPSS Statistics for Windows, version 22.0, Armonk, NY:  New diagnosis 35 (45.4)
IBM Corp.). Medians with ranges and percentages were used to de-
 Suspicion of CNS relapse 30 (39.0)
scribe continuous and categorical variables, respectively. Categorical
 End of treatment for CNS relapsea 12 (15.6)
variables were compared using chi-­square exact test. Student’s t test
a
Negative cerebrospinal fluid (CSF) sample after cytospin conventional cy-
was used to determine the relationship between continuous variables.
tomorphology (CCC) analysis.
The optimal WBC count from manual and automated CSF cell counts
that better predicted CNS involvement was then defined applying
receiver operating characteristic (ROC) curve analysis12,13 employing T A B L E   2   Results of 4 different cerebrospinal fluid (CSF) analysis
the MedCalc software V.16.8.4 (MedCalc Software, Ostend, Belgium), methods for detection of central nervous system (CNS) involvement
in 77 acute lymphoblastic leukemia (ALL) consecutive patients
with FCI results as the reference test. A P value < .05 was considered
according to clinical status at the time of the study
statistically significant.
Suspicion of End of therapy
New diagnosis CNS relapse for CNS relapse

3 | RESULTS n (%) 35 (45.4) 30 (39) 12 (15.6)

a
FCI 2 12 2
CSF samples from 77 patients were collected whose salient charac- CCC a
0 5 0
teristics are shown in Table 1. From these samples, 42 (54.5%) were ACCb 6 13 5
from male and 35 (45.5%) from female patients. Median age was MCC b
5 8 4
17 years (2-­61); ALL immunophenotype was B cell in 72 (93.5%) and
AAC, automated cell count; CCC, cytospin conventional cytomorphology;
T cell in 5 (6.5%) samples. There were 50 (64.9%) CSF samples from
FCI, flow cytometry immunophenotyping; MCC, manual cell count.
high risk and 27 (35.1%) from standard risk patients. Status by CCC a
Lymphoblasts.
b
at initial ALL diagnosis was CNS1 for all included individuals. Thirty-­ White blood cells (WBC).
five samples (45.4%) were from newly diagnosed individuals, 30 (39%)
belonged to patients in whom relapse was suspected, and 12 (15.6%)
CSF specimens were obtained from patients who completed treat- specificity, and predictive values for the tests were determined adopt-
ment for ALL relapse (Table 2). ing FCI results as the reference test and are shown in Table 4. Risk
From the whole cohort, FCI detected cells with lymphoblast anti- group had no influence on results of CCC and FCI (P = .811 and 0.719,
genic make-­up in 16/77 (20.8%) compared to 5/77 (6.5%) after cyto- respectively).
spin CC examination; MCC found more than 5 WBC in 17/77 (22.1%)
samples, and ACC was positive in 24/77 (31.2%) CSF specimens;
3.1 | Receiver operating characteristic curve analysis
comparison among the 4 methods for diagnosing CNS involvement in
ALL is shown in Table 3. FCI detected lymphoblasts in the CSF of 12 Median of manual WBC count in patients with CNS involvement
patients with symptoms suggestive of CNS relapse; only 5 (41.7%) of detected by FCI was 3.75 (0-­1280). In this regard, the optimal WBC
these samples were positive by CCC (P = .003). MCC that best correlated with CNS involvement according to FCI
Compared to FCI efficiency for detecting CNS involvement, the was 3.75 cells/μL, with a sensitivity of 50.0%, specificity of 85.2%,
test with higher sensitivity in our group was ACC at 71.4%, whereas an area under the curve (AUC) of 0.666, and a Youden index of
the test with higher specificity was CCC, with 100%. Sensitivity, 0.419. Median of automated CSF WBC count in patients with CNS
 |
4       JAIME-­PÉREZ et al.

T A B L E   3   Comparison of central
MCC+ MCC-­ ACC+ ACC-­ CCC+ CCC-­
nervous system (CNS) results for test
FCI positive 8 8 10 6 5 11 methods used to study cerebrospinal fluid
FCI negative 9 52 14 47 0 61 (CSF) samples of 77 acute lymphoblastic
leukemia (ALL) patients of all ages and at
P value .002 .002 <.001
different clinical stages
Flow cytometry immunophenotyping (FCI) was adopted as the reference test.
ACC, automated cell count; CCC, cytospin conventional cytomorphology; FCI, flow cytometry im-
munophenotyping; MCC, manual cell count.

T A B L E   4   Sensitivity, specificity, and predictive values for manual
(MCC) and automated (ACC) detection of white blood cells (WBC), or
lymphoblasts (cytospin conventional cytomorphology, CCC) in the
cerebrospinal fluid (CSF) of 77 ALL patients
MCC ACC CCC
studies;3 therapeutic implications of this occult CSF involvement
remain to be established by prospective studies with appropriate
length of follow-­up and randomization. It is important to under-
score that although sensitivity of FCI is considerably higher than
cytospin CC,14-16 this method has not been standardized, positive

Sensitivity (%) 50.0 62.5 31.2
Specificity (%) 85.2 70.5 100
PPV (%) 47.0 41.6 100
NPV (%) 86.6 88.6 84.7
Flow cytometry immunophenotyping (FCI) result in the same CSF sample
was used for comparison
NPV, negative predictive value; PPV, positive predictive value.
involvement detected by FCI was 5 (0-­1578). For automated cell
analyzer Sysmex XT-­4000i, the optimal WBC cutoff was 4.5 cells/
μL, with sensitivity of 62.5%, specificity of 70.5%, AUC of 0.676,
and Youden index of 0.41.
In a secondary analysis, patients with CNS involvement had a me-
dian of 28.5 (0.15-­97) lymphoblasts/μL identified by FCI. From the
16 FCI-­positive samples, the leukemic blast count per microliter was
greater in those 5 samples with a positive cytospin CC, with a mean
of 63.02 ± 43.08 vs 15.02 ± 18.28 in 11 with negative CCC (P = .019).
WBC count was greater in 10 CSF samples positive for both, FCI and
ACC, with a mean of 34.40 ± 38.40 vs 14.71 ± 6.90 in the remaining 4
detected only by FCI (P = .008), and also greater in 8 positive samples
for both FCI and MCC, with a mean of 35.49 ± 41.08 vs 19.20 ± 19.27
in the remaining 6 detected by FCI only (P = .389).
4 |  DISCUSSION
We report the results of a prospective study including 77 con-
secutive ALL patients from whom CSF samples were studied for
detection of WBC/lymphoblasts by 4 methods in different clini-
cal stages: at diagnosis, in the presence of symptoms suggesting
CNS relapse and after completing scheduled treatment for CNS
relapse with a negative control cytospin CC. The standard method
for detecting CNS involvement in ALL patients is cytomorphology
examination of a cytospin smear of CSF; our findings confirm stud-
ies concluding that this method has poor sensitivity compared to
FCI.2,3,11 Conventional cytospin cytomorphology detects CNS dis-
ease in up to 10% of patients at diagnosis of ALL, and this percent-
age is significantly lower than almost 30% recently reported by FCI
cutoff for CNS involvement is arbitrary, as recently underscored,10
its use varies among institutions, being mostly performed on high-­
income populations due to its considerable cost, and its clinical
meaning after detecting CNS involvement as a risk factor for re-
lapse remains largely inconclusive.9,17-19 In this respect, there are
studies supporting that CNS involvement detected only by FCI is
associated with a higher incidence of relapse and lower overall sur-
vival2,11 suggesting that patients with positive findings with this
method could benefit from additional intensified systemic and/
or CNS-­directed therapy; it is relevant to point out that no cur-
rent clinical guideline conclusively supports this intensified treat-
ment approach. On the other hand, there is also evidence that CNS
involvement detected only by FCI at diagnosis and negative by
cytospin CC does not lead to relapse; in a study in children from
Sweden, including 214 patients at diagnosis of ALL, CSF involve-
ment was positive by both methods, FCI and cytospin CC, in twenty
patients, an additional seventeen were detected only by FCI; none
of these 17 children relapsed to the CNS during long-­term follow-
­up.2 Thus, the use of FCI of CSF as an additional method to assess
CNS involvement in ALL and to guide therapeutic decisions based
on its results remains a controversial subject and its meaning needs
to be interpreted taking into account individual patient character-
istics as well as specific population variables, including proportion
of high-­risk patients, intensity of both, systemic and intrathecal
therapy, and long-­term results of treatment protocols.
In our group, leukemic blast count by FCI was significantly greater
in samples with a positive cytospin CC than in those positive only by
FCI, suggesting that operator-­dependent morphological conventional
cytospin evaluation, even if carried out by experimented observers,
performs suboptimally detecting involvement of the CNS in ALL
patients.
Interestingly, from 30 patients with clinical suspicion of CNS re-
lapse in our group, twelve, almost 40%, had lymphoblasts detected
by immunophenotyping, whereas only 5, less than 50%, were positive
by conventional cytospin examination. Although the reduced number
of studied samples restrict these conclusions, it is interesting that the
ROC cutoff values for manual and automated cell counts are close to
those commonly used to define CNS involvement (>5 white blood
cells per microliter and a positive cytospin).
JAIME-­PÉREZ et al.
Principal limitations in this proof-­of-­concept study are a relatively
small sample size and heterogeneity in age groups and clinical stages;
however, the hypothesis of FCI having higher sensitivity than CCC for
detecting the presence of lymphoblasts in the CSF of ALL patients at
different clinical points was confirmed; the meaning of this finding in
therapeutic decision-­making among the different clinical situations,
however, is yet to be established.
In addition to differences in technical performance, cost is an im-
portant consideration when assessing test efficiency, and this is more
important in financially restricted health environments. In our labo-
ratory, MC+CCC for CSF analysis in patients with ALL costs $12.50
USD; ACC of WBC performed in the Sysmex XT-­4000i amounts to
$6 USD per CSF sample. Both methods are quite cheaper than FC im-
munophenotyping, which costs $287 dollars per sample at our public
institution. In this respect, ACC had more than twice the sensitivity
for detecting patients with WBC in the CNS than CCC, although ACC
is considerably limited by the fact that it cannot discriminate lympho-
blasts from other WBC, including normal lymphocytes and mono-
cytes; thus, ACC could serve as an inexpensive rapid screen test to
decide which samples should be further studied by FCI. However, it
is important to note that, if this had been the case, 6 (37.5%) ACC-­
negative patients but FCI positive would have been missed.
In conclusion, FCI of the CSF had higher sensitivity than cytospin CC
in detecting CNS involvement in ALL patients at different clinical sce-
narios and it could complement CCC performance. Studies appropriately
powered including long-­term follow-­up are needed to elucidate the clini-
cal meaning of low-­grade CNS involvement detected only by FCI.
CO NFLI CT OF I NTERE S T
The authors declare no conflict of interest.
AUT HOR CONTRI BUTI O N
José Carlos Jaime-­Pérez conceptualized and designed the study,
drafted the initial manuscript, and approved the final manuscript as
submitted. María Fernanda Borrego-­López carried out the initial anal-
yses, reviewed the manuscript, and approved the final manuscript as
submitted. Raúl Alberto Jiménez-­Castillo designed the data collection
instruments, supervised data collection, and approved the final manu-
script as submitted. Nereida Méndez-­Ramírez participated in study
design, critically reviewed the manuscript, and approved the final
manuscript as submitted. Rosario Salazar-­Riojas revised the manu-
script, coordinated data collection, and approved the final manuscript
as submitted. Lucía Teresa Fernández participated in study design and
approved the final manuscript as submitted. David Gómez-­Almaguer
critically reviewed the manuscript and approved the final manuscript
as submitted.
O RCI D
J. C. Jaime-Pérez  http://orcid.org/0000-0001-6804-9095
|
      5
REFERENCES
1. Lazarus HM, Richards SM, Chopra R, et al. Central nervous system in-
volvement in adult acute lymphoblastic leukemia at diagnosis: results
from the international ALL trial MRC UKALL XII/ECOG E2993. Blood.
2006;108:465‐472.
2. Ranta S, Nilsson F, Harila-Saari A, et al. Detection of central nervous
system involvement in childhood acute lymphoblastic leukemia by cy-
tomorphology and flow cytometry of the cerebrospinal fluid. Pediatr
Blood Cancer. 2015;62:951‐956.
3. Levinsen M, Marquart HV, Groth-Pedersen L, et  al. Leukemic
blasts are present at low levels in spinal fluid in one-­third of child-
hood acute lymphoblastic leukemia cases. Pediatr Blood Cancer.
2016;63:1935‐1942.
4. Jabbour E, Thomas D, Cortes J, Kantarjian HM, O’Brien S.
Central nervous system prophylaxis in adults with acute lym-
phoblastic leukemia: current and emerging therapies. Cancer.
2010;116:2290‐2300.
5. Lohajaroensub S, Sakoonwatanyoo P, Sakunthaworn S, Pichanun D.
Comparison of body fluid cell counting between automate hema-
tology analyzer sysmex XT-­4000i and manual microscopic method.
Vajira Med J. 2016;59:21.
6. Glantz MJ, Cole BF, Glantz LK, et  al. Cerebrospinal fluid cytology
in patients with cancer: minimizing false-­negative results. Cancer.
1998;82:733‐739.
7. Pui CH, Thiel E. Central nervous system disease in hematologic malig-
nancies: historical perspective and practical applications. Semin Oncol.
2009;36(4 Suppl 2):S2‐S16.
8. de Graaf MT, de Jongste AH, Kraan J, Boonstra JG, Sillevis Smitt PA,
Gratama JW. Flow cytometric characterization of cerebrospinal fluid
cells. Cytometry B Clin Cytom. 2011;80:271‐281.
9. Burger B, Zimmermann M, Mann G, et  al. Diagnostic cerebrospinal
fluid examination in children with acute lymphoblastic leukemia: sig-
nificance of low leukocyte counts with blasts or traumatic lumbar
puncture. J Clin Oncol. 2003;21:184‐188.
10. Dass J, Dayama A, Mishra PC, et  al. Higher rate of central nervous
system involvement by flow cytometry than morphology in acute
lymphoblastic leukemia. Int J Lab Hematol. 2017;39:546‐551.
11. Martinez-Laperche C, Gomez-Garcia AM, Lassaletta A, et  al.
Detection of occult cerebrospinal fluid involvement during
­maintenance therapy identifies a group of children with acute lym-
phoblastic leukemia at high risk for relapse. Am J Hematol. 2013;88:
359‐364.
12. Hanley JA, McNeil BJ. The meaning and use of the area under a re-
ceiver operating characteristic (ROC) curve. Radiology. 1982;143:
29‐36.
13. Zweig MH, Campbell G. Receiver-­operating characteristic (ROC)
plots: a fundamental evaluation tool in clinical medicine. Clin Chem.
1993;39:561‐577.
14. Subira D, Castanon S, Roman A, et al. Flow cytometry and the study
of central nervous disease in patients with acute leukaemia. Br J
Haematol. 2001;112:381‐384.
15. Hegde U, Filie A, Little RF, et al. High incidence of occult leptome-
ningeal disease detected by flow cytometry in newly diagnosed
aggressive B-­cell lymphomas at risk for central nervous system
involvement: the role of flow cytometry versus cytology. Blood.
2005;105:496‐502.
16. Galati D, Di Noto R, Del Vecchio L. Diagnostic strategies to investigate
cerebrospinal fluid involvement in haematological malignancies. Leuk
Res. 2013;37:231‐237.
17. Smith M, Arthur D, Camitta B, et al. Uniform approach to risk classifi-
cation and treatment assignment for children with acute lymphoblas-
tic leukemia. J Clin Oncol. 1996;14:18‐24.
18. Pui CH, Boyett JM, Rivera GK, et  al. Long-­term results of Total
Therapy studies 11, 12 and 13A for childhood acute lymphoblastic
 |
6       JAIME-­PÉREZ et al.

leukemia at St Jude Children’s Research Hospital. Leukemia.

2000;14:2286‐2294.
1 9. Biojone E, Queiroz Rde P, Valera ET, et  al. Minimal residual dis-
ease in cerebrospinal fluid at diagnosis: a more intensive treat-
ment protocol was able to eliminate the adverse prognosis in
children with acute lymphoblastic leukemia. Leuk Lymphoma.
2012;53:89‐95.
How to cite this article: Jaime-Pérez JC, Borrego-López MF,
Jiménez-Castillo RA, et al. Comparison of conventional
cytomorphology, flow cytometry immunophenotyping, and
automated cell counting of CSF for detection of CNS
involvement in acute lymphoblastic leukemia. Int J Lab Hem.
2017;00:1–6. https://doi.org/10.1111/ijlh.12760