Professional Documents
Culture Documents
DOI: 10.1111/ijlh.12935
ORIGINAL ARTICLE
1
Clinical Patology Laboratory, A.O.R.N. “G.
Rummo” di Benevento, Benevento, Italy Abstract
2
Department of Laboratory Medicine Introduction: Acute promyelocytic leukemia (APL) is a type of acute myeloid leuke-
EOLAB, Ente Ospedaliero Cantonale,
mia (AML) with a life-threatening coagulopathy. Once it is suspected, ATRA should be
Bellinzona, Switzerland
3 started. Appreciation of APL details is critical, but an experienced hematopathologist
Department of Cellular Biotechnologies
and Hematology, Sapienza University, Rome, may not be available. We developed an algorithm, based on the parameters gener-
Italy
ated by automated blood cell counter ADVIA 2120i Siemens that can aid the diagno-
4
Department of Molecular
Medicine, Policlinico Umberto I, Sapienza sis of APL.
University, Rome, Italy Methods: All parameters in the algorithm were selected on the bases of the patho-
5
Department of Laboratory physiology of the APL and the analyzer’s technology. We used c1 cutoff:
Medicine, Azienda Sanitaria Locale Regione
Liguria, Lavagna, Italy PLT < 150 × 103; % Mono<10; % LUC<4; % hyperchromic cells > 2; %saturated cells ≥
6
Department of Laboratory Medicine, 1; % blasts > 4. Satisfying at least five of six cutoffs, we obtained an “APL criteria”. It
Civitanova Marche, Italy
was tested on 247.209 raw-data from routine and emergency samples and on 124
Correspondence raw-data from APL. We performed a multiparametric analyses and obtained defini-
Dr. Vincenzo Rocco, Clinical Patology
tive cutoffs (c2). It was validated on a new group of 51.002 raw-data from routine and
Laboratory—A.O.R.N. “G.Rummo” di
Benevento, Benevento, Italy. emergency. Finally, it was tested on AML and ALL, MDS, MPN, oncologic samples,
Email: vincenzorocco1@virgilio.it
and hemolytic and megaloblastic anemia.
Results: The algorithm provided a high sensitivity (86.30%) and high specificity
(99.83%) in APL with high or normal number of WBC and satisfactory results in APL
with cytopenia (80%). The specificity was very high also in groups of hematological
and nonhematological diseases. It can be used as an “APL criteria” to alert patholo-
gists of the possible presence of APL. It together with instrumental and classical mor-
phology may allow to reduce the time of diagnosis with further reduction in early
death.
1 | I NTRO D U C TI O N (ATRA) in 1995, the disease has become the most curable subtype
of adult AML.3-5
Acute promyelocytic leukemia (APL) is a subtype of acute myeloid Hence, once the APL is suspected on the basis of clinical findings
leukemia (AML) characterized by a specific balanced reciprocal and the peripheral blood smear, bone marrow aspirate should be
translocation t (15;17), which fuses the PML gene on chromosome highly recommended, but even without waiting for its examination,
15 to the RAR-α (retinoic acid receptor-α) gene on chromosome 17.1,2 and before the diagnosis is confirmed by cytogenetic or molecular
Most patients exhibit a life-
threatening coagulopathy. Since studies, ATRA should be started at the standard dose, to promptly
the introduction of the differentiating agent all-trans retinoic acid resolve the coagulopathy.6,7
Int J Lab Hematol. 2018;1–9. © 2018 John Wiley & Sons Ltd | 1
wileyonlinelibrary.com/journal/ijlh
|
2 ROCCO et al.
Morphologically, APL is not a homogeneous entity; the classical or kidney-shaped nuclei with very minute granules in the cytoplasm
hypergranular APL represents the great majority of all APL, and the that are not readily perceptible on light microscopy.8-10
microgranular variant accounts for about 15%-20% of all APL. In the Appreciation of these morphological details is critical; the short-
classical hypergranular APL, there are abnormal promyelocytes with age of skilled and experienced morphologists in some setting means
an abundant cytoplasm packed with large azurophilic granules. Auer that slides have to be transported to specialized laboratories for re-
bodies are commonly found in many cells. The microgranular vari- view.6 This leads to unnecessary diagnostic delays and may compro-
ant is characterized by the presence of leukemic cells with bilobed mise patient’s outcome.
F I G U R E 1 Caracteristic pattern of RBC V/HC, Perox and BASO channel from ADVIA 2120i SIEMENS in A, normal sample, in B, APL
with hight WBC count, in C, cytopenic APL sample. %Bl, Blasts percentage; %Iper, Hyperchromic RBC; Ba, Basophils; E, Eosinophils; L,
Lymphocytes; LUC, Large Unstained Cells; M, Monocytes; N, Neutrophils; Perox % Sat, Percentage of Cells in perox saturation [Colour
figure can be viewed at wileyonlinelibrary.com]
ROCCO et al. |
3
Many studies reported that parameters generated by auto- peroxidase, called Large Unstained Cells (LUC). The basophil/
mated blood cell counters have a possible role in aiding the diagnosis nuclear lobularity channel (ND) is used to derive nuclear density.
and classification of acute leukemia.11-13 For example, Peroxidase Phthalic acid strips the cytoplasm from white blood cells except
Activity and Nuclear Density Analysis (PANDA) classification, based basophils. The WBCs pass through a laser flow cell and two-angle
on the morphology of cell clusters on ADVIA 120 and 2120i, has light scatter is then used to determine cell size and nuclear density
been proposed and confirmed as a useful tool for a premicroscopic that are detected and recorded on the basophil cytogram in which
assessment of leukemic samples.10,14,15 In this context, APL provides the x-a xis reflects the nuclear complexity and the y-a xis reflects
the most characteristic pattern among all leukemias as shown by cell size.
D’Onofrio, Gibbs, and Pillay, because the MPO activity in promy- In the basophil channel, the analyzer also provides a % value of
elocytes is maximum and they are arranged on the extreme right presumed blasts. In fact, the chromatin structure of blast cells is not
of the perox channel and straddling the population of eosinophils very dense and this feature causes a low signal of high-angle scatter-
(Figure 1B-C).16-20 ing resulting in a location of the immature nuclei in the left proximal
Another study led to design of a diagnostic algorithm for APL part of the head of mononuclear (MN) cells. 21
using as parameters the percentage of large unstained cells (LUC)
and delta neutrophil index (DNI), but this work has been done on
2.1.2 | RBC count and classification
bone marrow aspiration fluid. 21
On the basis of these premises, and of the consideration that the Erythrocytes are counted and classified on the bases of volume
evaluation of complete blood count (CBC) is the first test that the pa- and hemoglobin concentration. By applying Mie’s theory, 22 a soft-
tient performs, often in emergency situations (access to Emergency ware is able to produce a single cytogram allowing to quantify the
Departments for hemorrhage), the purpose of our study was to build different erythrocyte populations in a sample of peripheral blood
an algorithm that allows to produce a “APL criteria,” on ADVIA 2120i respect to their volume and hemoglobin concentration, provid-
(Siemens), with high sensitivity and very high specificity, in order to ing, among others, the percentage of hyperchromatic red blood
alert biotechnologists and pathologists of the possible presence of cells (% Hyper) presumably due to spherocytes and microspheres.
the disease and not to burden on the loads working with the produc- (Figure 1A).
tion of a high number of false positives.
2.1.3 | PLT count
2 | M ATE R I A L A N D M E TH O DS The ADVIA 2120i uses a two-dimensional platelet analysis. This
method is based on the simultaneous measurement of laser light
scattered at two different angles. Both measurements are converted
2.1 | Analyzer technology
into volume and refractive index. A cytogram is produced, where
The Siemens Advia 2120i (Siemens Healthcare Diagnostics, platelets are identified in the region corresponding to a volume of
Deerfield, Illinois, USA) is an automated hematology analyzer that 1-60 fL and refractive index of 1.35-1.40. 21,23-25
measures the total and differential WBC count using principles of For each analyzed sample, the ADVIA 2120i (as well as its prede-
flow cytometry and a combination of reactions that occur within cessor ADVIA120) automatically saves all data, related to cell counts
the peroxidase and the basophil/nuclear lobularity channels. Cluster and their respective positional parameters, in a single file, called
analysis of the cells within each channel is used to generate a cyto- raw-data. These data are tabulated and can subsequently be viewed
gram and the pattern that emerges has been shown to assist with the and processed using any spreadsheet such as Microsoft Excel®.
10,14-18
subtyping of hematological malignancies.
TA B L E 1 Raw-data and number of cases about groups of At this point, we kept the cutoff constant for five of the six pa-
hematological and nonhematological diseases rameters, varying time to time the sixth one and calculating sensitiv-
Raw-data ity and specificity; those with higher sensitivity and specificity have
Diseases (n = 800) Cases (n = 650) been used as c2 cutoffs.
Then, to overcome the possible bias due to different calibrators
AML no APL 142 142
(Siemens ADVIA 2120i SETpoint™ Calibrator) and different analyz-
ALL 36 36
ers, standardization used in other laboratories, the specificity of the
MDS 90 90
algorithm for the APL-Flag, was tested on further 51.002 raw-data
MPN 47 47
of peripheral blood samples (from routine and emergency) from two
Megaloblastic anemia 30 30
laboratories participating in the study.
Acute and chronic hemolytic 39 39 To verify the specificity of this algorithm, with respect to inter-
anemia
fering factors deriving from particular conditions, it has been tested
Cytopenic cancer patients 34 34
on groups of patients affected from hematological and nonhemato-
Cancer patients with neutro- 134 134
logical diseases.
philic leukocytosis (by Growth
factors therapy)
APL 124 49 2.3 | Groups in study
Typical 112 45
Atypical 12 4
1. A total of 247.209 raw-data of CBC from routine (109.159)
and emergency (138.050) APL-negative
4. Increased % hyperchromic RBC: because the DIC, which is almost
always present in the APL, causes an increased percentage of This evaluation was necessary for the construction of the algo-
schizocytes and consequently of microspherocytes26,27; rithm. In fact, a too low specificity, and therefore with a high number of
5. Increased % Perox saturated cells: because the promyelocytic false positives, would lead to microscopic control of too many useless
blasts, due to their very high peroxidase content, are placed be- samples.
tween the saturated cells, leading to an increase in this It cannot be justifiable in these cases that almost all false posi-
parameter; tives are related to blood samples that should be lead to microscopic
6. Increased % blasts: because, due to the reduced chromatin den- review.
sity of the APL cells, a portion higher than 4% exceeds the prede- The actions resulting from the suspicion of APL are so peculiar
termined threshold and this causes these cells to be counted as and urgent that many false positives could lead to the renunciation
blasts by the analyzer. of the use of the flag;
Considering the normal values of the above-mentioned parame- 2. A total of 51.002 raw-data of CBC from routine and emergency
ters, supplied in default by the analyzer, and their characteristic alter- assessed in a single group, coming from two others
ations, we have chosen, for each of them, the following cutoff (cutoffs laboratories.
c1) to insert in the algorithm: 3. Hematological and nonhematological diseases (Table 1) with cy-
tometric and morphological characteristics, which could produce
• PLT < 150 × 103/μL; false positives due to the possible presence of:
• % Mono < 10; a. Cells with high peroxidase content;
• % LUC < 4; b. Reduced LUC number;
• % hyperchromic cells > 2; c. Real or spurious increased blasts percentage;
• % saturated cells ≥ 1; d. Thrombocytopenia.
• % blasts > 4. e. Increased percentage of spherocytes.
f. Monocitopenia;
Firstly, for each of the c1 cutoffs, the statistical significance was g. Schizocytes and consequently of microspherocytes26-28
evaluated by comparing APL vs normal samples. 4. 127 raw-data from 127 donors group (normal samples)
Subsequently, this algorithm was built on 247.209 sample files
(raw-data), deriving from the emergency and routine samples pro- All 124 raw-data of 49 cases of APL (112 typical and 12 atypical)
vided by clinical laboratory of AORN “Rummo” (pilot laboratory). confirmed according to the WHO guidelines.29
With such cutoffs (c1), the algorithm A1 was constructed; the al- The files were collected in the years 2014-17 in five laboratories
gorithm provided that at least five of six criteria must be verified in involved in the study.
order to trigger the “APL criterion.” With these c1 cutoffs, we have Clinical requests (CBC with differential formula, morphology,
obtained an initial sensitivity and specificity value. and cytofluorimetry) related to each case in the study were made
ROCCO et al. 5|
by the clinicians as an integral part of the diagnostic process . No the percentage of hyperchromic cells, as well as for the percentage
samples were requested from the patient for this study. of cells in saturation and the percentage of blasts, we found a sta-
tistically significant difference (Table 2). Monocyte count and per-
centage of LUC, although in the normal range, showed a statistically
2.4 | Statistical analysis
significant trend (P < 0.05) at the lower limits (Figure 2 and Table 2).
We used raw-data provided by the ADVIA2120i (Siemens) for
each sample, to obtain a full list with all the parameters provided
3.2 | Diagnostic accuracy for parameters
by analyzer. Statistical analysis was performed using Microsoft
in the study
Excel (Microsoft, Redmond, WA, USA) and MedCalc software ver-
sion 11.4.2.0 (MedCalc, Ghent, Belgium). To test the distribution Considering the statistically significant difference between normal
of the data, we used the D’Agostino-Pearson test because, for and APL samples found for each individual parameter, it was decided
each group, we observed a P-value < 0.05 (data not fit a normal to evaluate their diagnostic accuracy using cutoffs “c1” tested on
distribution); data were presented as median, minimum and maxi- the groups routine and emergency APL-negative and APL-positive
mum, and interquartile range (IRQ), and a nonparametric method (Table 3).
was used . To display variation in groups, Box&Whishers distribu- Except for the % Iper, the “c1” cutoff showed a good sensitivity
tion graphs were used; Mann-W hitney test was used to analyze (>80%) for all parameters. About specificity, a low value is observed
association (a “P” value lower than 0.05 was considered statisti- for monocytes and LUC parameters (Table 3).
cally significant). Then, these parameters with “c1” cutoffs were implemented in
The main end-point of the study was the construction of an al- an algorithm A1, according to which at least five of six criteria had
gorithm to predict the APL, so we used a group to build algorithm to be satisfied. This allowed to identify correctly the 89.16% of APL
(247.209 samples from routine and emergency), and a group to vali- raw-data and the 99.65% of APL-negative samples.
date it (51.002 samples from routine and emergency). Thus, to improve these cutoffs, we analyzed the individual pa-
We used a univariate descriptive statistical analysis to compare rameters as described in the paragraph “Parameters and construc-
groups and parameters individually. tion of the algorithm” and we obtained cutoffs “c2” (Table 3).
The diagnostic accuracy of all cutoff points and algorithm was Through these “c2” cutoffs, it was possible to obtain a sensitiv-
determined by calculating sensitivity and specificity tested on ity of 87.75% (43/49) patients at first diagnosis and a sensitivity of
the groups APL-positive and on total routine and emergency APL- 86.30% (107/124) for all raw-data of patients with APL; the specific-
negative, respectively, toward individual groups of hematological ity was 99.83% (246.789/247.209) (VPP 18.77%, VPN 99.99%, LR+
and nonhematological diseases ; positive and negative predictive 507, LR− 0.002).
values (PPV and NPV) and likelihood ratios (LR+, LR−) have also been In particular, the algorithm correctly classified 93.4% (14 of 15
calculated. samples) of APL with normal count of WBC, 100% of the samples
with leukocytosis (42 of 42 samples) and surprisingly 80.6% of the
cytopenic APL (54 of 67 samples) (Figure 1B-C and Table 4).
3 | R E S U LT S
Analyzing the false positive (FP) from the total groups of rou-
tine and emergency APL-negative (n = 247.209), it was found that
3.1 | Normal samples vs APL: Comparative analysis
14.76% (n = 62) of them was represented by Megaloblastic Anemias
The comparison between normal samples and APL samples, re- and transfused Megaloblastic Anemias, 9.52% (n = 40) by oncologic
spect to the selected and above-mentioned parameters, gave the samples, and 57.38% (n = 241) by “reactive” samples. As for the
following results; as expected, both for the platelet count and for 18.34% (n = 77) remaining samples, it was represented by “reactive”
F I G U R E 2 Box & Whiskers plot of normal samples vs APL group. % LUC, percentage of Large Unstained Cells; % Mono, percentage
of monocytes; hyper_pcnt, percentage of hyperchromic cells; pcnt_blast (%), percentage of cells to the left of the BASO channel blast
threshold; pcnt_hight absorp (%), percentage of perox saturated cells; PLT, Platelets [Colour figure can be viewed at wileyonlinelibrary.com]
TA B L E 4 Evaluation of specificity and sensitivity of the A2 algorithm in APL group and hematological and nonhematological diseases,
respectively
available. It was found: one false positive both in the MDS group In the Megaloblastic Anemia group, the only false-p ositive
and in Hemolytic Anemias in which five of six criteria were verified sample met five of six required criteria (except for the % Iper).
(except for the % blasts). (Table 4)
|
8 ROCCO et al.
Johannesburg March 2015. Link to: https://pdfs.semanticscholar. 26. Zini G, D’Onofrio G, Briggs C, et al. ICSH recomandations for iden-
org/21bd/171eba483e26184cee28929f514aecc79f78.pdf tification, diagnostic value and quantitation of schizocytes. Int J Lab
20. D’Onofrio G, Zini G. Automated cytochemistry of acute promyelo- Hematol. 2012;34:107‐116.
cytic leukemia: there’s more than numbers. Int J Lab Hematol. 2015 27. Bain JB. Blood cell morphology in health and disease. Dacie and
Feb;37(1):137‐138. Lewis Practical Hematology, Vol. 5. Philadelphia, PA:Churchill
21. Jang MJ, Choi HW, Lee SY, et al. Application of bone marrow sam- Livingstone; 2006:79‐113.
ples for discrimination of acute promyelocytic leukemia from other 28. Bain BJ. Schistocytes in megaloblastic anemia. Am J Hematol.
types of acute leukemia using the routine automated hematology 2010;85(8):599.
analyzer. Int J Lab Hematol. 2014;36:531‐540. 29. Swerdlow SH, Campo E, Harris N, et al. WHO Classification of
22. ADVIA. 2120: White blood cell technology. Available from: www. Tumours of Haematopoietic and Lymphoid Tissues, 4th edn. Lyon,
medical.siemens.com France: IARC; 2008.
23. Mie G. Beiträge zur Optik trüber Medien, speziell kolloidaler
Metallösungen”. Ann Phys. 1908;330(3):377‐445.
24. Fisher G, Zelmanovic D, Shapiro P, Greenbaum A, Kunicka JE.
How to cite this article: Rocco V, Castelli C, Fumi M, et al.
Evaluation of the Two Dimensional Platelet Method on the ADVIA
The diagnostic use of ADVIA 2120i Siemens and an “APL
120TM Hematology System. In First International Bayer Diagnostics
Central Laboratory Symposium. Wilton, Conn: Chase Medical criteria” can help to reduce the rate of early death in the APL.
Publishers; 1997:46. Int J Lab Hematol. 2018;00:1–9. https://doi.org/10.1111/
25. Giacomini A, Legovini P, Antico F, et al. Evaluation of platelet ijlh.12935
analysis on the ADVIA 120 hematology system. Lab Hematol.
2001;7:180‐185.