Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders

J. Comp. Path. 2014, Vol. -, 1e13

Available online at www.sciencedirect.com

ScienceDirect www.elsevier.com/locate/jcpa

SPONTANEOUSLY ARISING DISEASE

Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders E. Tan*, A. C. G. Abrams-Ogg†, A. Defarges† and D. Bienzle* * Department of Pathobiology and † Department of Clinical Studies, University of Guelph, 50 Stone Road East, Guelph, Ontario N1G 2W1, Canada

Summary Automated analysis of bone marrow (BM) aspirates is a useful ‘pre-microscopical’ screen to identify hypocellular samples and those with potentially abnormal cells. In order to determine whether automated analysis could also be used to identify haemopoietic abnormalities, EDTA-anticoagulated BM aspirates from 43 dogs were analysed using the Advia 2120 instrument. Corresponding Wright-stained BM smears were evaluated microscopically to determine smear quality, cell composition and 500-cell differential counts, and correlation to automated analysis parameters was computed. Leucocyte cytograms generated by the automated analyzer were scrutinized and compared with those of ‘normal’ BM. Twenty-three neoplastic and 20 nonneoplastic samples were analysed, including samples from 10 cases of acute myeloid leukaemia, four cases of acute lymphocytic leukaemia, four cases of chronic lymphocytic leukaemia, one case of chronic neutrophilic leukaemia, three cases of multiple myeloma, one case of myelodysplastic syndrome, five cases of nonregenerative immune-mediated haemolytic anaemia, one case of immune-mediated neutropenia, three cases of immune-mediated thrombocytopenia, six cases of inflammatory disease, three samples with myelotoxicity and two samples analysed for staging of neoplasia. Automated white blood cell (WBC) counts correlated significantly with smear cellularity, particle cellularity and particle number. There was a significant difference in WBC counts of samples with insufficient versus sufficient particles. Significant correlations between Advia percent neutrophils and microscopical determination of marrow segmented neutrophils/neutrophilic granulocyte reserve, Advia percent lymphocytes and microscopical determination of lymphocytes/rubricytes, Advia percent large unstained cells and microscopical determination of myeloblasts/promyelocytes and between Advia percent eosinophils and manual determination of eosinophils were identified. This suggested that Advia WBC counts may be used to approximate BM sample quality and that Advia differential counts may predict marrow granulocyte reserve and lymphocyte/rubricyte stores. Distinct and consistent alterations in cytogram patterns were observed in cases of acute leukaemia, but were less obvious in chronic leukaemia. Complete automated BM analysis was performed in approximately 2 min, while staining and coverslipping of BM slides required approximately 30 min. Hence, although automated analysis should not supplant microscopical evaluation of BM, it can provide useful ancillary information in a short time and flag potentially inadequate or abnormal samples. Ó 2014 Elsevier Ltd. All rights reserved. Keywords: Advia 2120; haematology analyzer; haemopoiesis; leukaemia

Introduction Bone marrow (BM) assessment is an important component of diagnostic and toxicological investigation of haematological disorders. In clinical settings, BM aspirate and core biopsy are indicated when Correspondence to: E. Tan (e-mail: [email protected]). 0021-9975/$ - see front matter http://dx.doi.org/10.1016/j.jcpa.2014.02.005

persistent or unexplained abnormalities in the peripheral blood are present, such as cytopenia, cytosis or presence of atypical cells, or for staging of neoplasia (Harvey, 2012). In addition to qualitative assessment of slides by a pathologist, differential counting of 200e1,000 cells is recommended, in particular for toxicological assessment (Provencher-Bollinger, 2004; Travlos, 2006; Riley et al., 2009). However, Ó 2014 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

E. Tan et al.

2

differential counting appears to be performed less frequently in diagnostic settings, possibly due to the labour-intensive and time-consuming nature of this procedure (Harvey, 2012). Automated haematological analysis of BM aspirates can provide valuable pre-microscopical information concerning the quality of samples, presence of cell types and the potential need for further analysis. For example, Fan et al. (1999) found that BM analysis with the Cell Dyn 4000 instrument compared with microscopical analysis was sensitive and specific for detection of granulocytic hyperplasia (86% and 88%, respectively), lymphocyte-predominance (97% and 92%), increased blast cell proportion (89% and 98%) and non-diagnostic samples (68% and 96%). Previously, we demonstrated the utility of automated analysis of BM aspirates from healthy dogs using the Advia 2120 haematology analyzer (Siemens Healthcare Diagnostics, Tarrytown, New York, USA; Tan et al., 2013). In that study, automated results were available within approximately 2 min of the sample arriving in the laboratory, while staining and coverslipping slides required approximately 30 min. Automated results correlated well with microscopical evaluation of BM smears and provided useful ancillary information regarding the potential diagnostic quality of samples and the adequacy of marrow granulocyte and rubricyte reserves. We now wished to determine whether automated analysis would prove similarly useful in the analysis of BM samples from dogs with haematological abnormalities. The specific aims of this study were: (1) to determine whether results of automated analysis correlated with microscopical assessments of a BM smear; and (2) to investigate whether Advia cytograms could provide useful information about haematological disorders.

Materials and Methods Bone Marrow Samples

BM samples were obtained from dogs admitted to the Ontario Veterinary College Health Sciences Centre (OVC-HSC) for investigation of haematological abnormalities (e.g. cytopenia or cytosis, atypical cells detected on blood smears) or staging of neoplasia. BM was aspirated from the manubrium, iliac crest or humerus as deemed appropriate by the attending clinician. BM fluid (approximately 0.5 ml) was aspirated into a syringe containing 0.6 ml saline with 1% sterile ethylenediaminetetraacetic acid (EDTA). One drop of the aspirate was placed on each of three to five glass slides, blood was removed by tilting the slides onto absorbent paper, smears were prepared

by squash preparation and stained twice with modified Wright’s stain using an automated stainer (Hema-TecÒ 2000, Siemens Healthcare Diagnostics, Tarrytown, New York, USA; Bienzle, 2012; Defarges et al., 2013). The remainder of the BM aspirate was placed into a 3 ml EDTA tube (Becton Dickinson, Franklin Lakes, New Jersey, USA) for automated analysis. BM smears and fluid were processed within 2 h of collection. Any clots apparent prior to automated analysis were removed manually with a pipette and if clots were multiple or large, the sample was excluded from automated analysis. Samples without detectable particles on microscopical analysis were also excluded from analysis. Bone Marrow Automated Analysis

The Advia 2120 is a multichannel, optical plus cytochemical, automated haematology analyzer that generates a five-part white blood cell (WBC) count and enumerates reticulocytes. The differential WBC count in the Advia is based on basophil/lobularity and peroxidase analyses in different channels. In the basophil/lobularity channel, cells are exposed to a reagent that lyses red blood cells (RBCs), platelets and the cytoplasm of most leucocytes. Basophil enumeration in this channel is based on cluster analysis of size (forward low-angle light scatter) versus nuclear lobularity, since human basophils are more resistant to lysis than other leucocytes (Gibbs, 2011). Basophils from dogs and cats do not appear to be lysis resistant and are not enumerated accurately with this method (Lillieh€o€ok and Tvedten, 2011). In the peroxidase channel, leucocytes are categorized by cluster analysis based on size (forward low-angle light scatter) versus cytoplasmic peroxidase content. This method discriminates five unique clusters in canine samples: neutrophils, eosinophils, monocytes, lymphocytes and large unstained cells (LUCs; Fig. 1A). This latter cell type refers to cells that are large, but lack peroxidase activity (PA), and may include blast cells of various lineages. Basophils are not enumerated separately in the Advia peroxidase channel, since they appear to cluster with lymphocytes or LUCs (Gibbs, 2011). The Advia 2120 reticulocyte concentration is derived from analysis of RNA content of isovolumetrically sphered RBCs through binding of oxazine 750 to RNA, which reduces light transmission in anucleate cells. BM aspirates were analysed in the Advia 2120 using the manufacturer’s haematology settings for dogs, software version 5.9 MS. Parameters collected included the following: total WBC counts (from the basophil/lobularity channel), results of the automated

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

Automated Bone Marrow Analysis in Dogs

3

Fig. 1. Advia basophil/lobularity (left) and peroxidase (right) cytograms for selected diagnostic categories. (A) Peripheral blood from a haematologically normal dog, WBC ¼ 8.9  109/l. Peroxidase cytogram shows clusters for (1) lymphocytes; (2) LUCs; (3) monocytes; (4) neutrophils; and (5) eosinophils. (B) BM aspirate of inadequate cellularity (no BM particles), WBC ¼ 16.5  109/l. The cytograms have similar density and distributions as those of peripheral blood, indicating predominance of peripheral blood with addition of fat particles (circle in basophil/lobularity and peroxidase channels), greater number of lysis-resistant cells (upper box on basophil/lobularity) and cells in LUC region, indicative of a moderate number of haematopoietic precursors. (C) BM sample from haematologically normal young dog, WBC ¼ 413  109/l. The cytograms are denser than those of peripheral blood, indicating high cellularity of sample. In contrast to peripheral blood, BM samples have a prominent cell cluster in the LUC region. Cell density in the neutrophil region likely corresponds to post-mitotic granulocytic cells. Blast and immature cells are likely to populate the lysisresistant region of the basophil/lobularity cytogram. (D) Granulocytic hyperplasia and (E) CNL. Note the similarity between these cytograms and those of normal BM. (F) ALL with absence of normal haemopoietic cells in neutrophil and eosinophil regions. (G) AML. Note the dense cell cluster extending from the LUC region to the upper right corner of the peroxidase cytogram (circle), indicating large cells with a continuum of peroxidase activity. (H) Large granule lymphocyte CLL involving BM. There is an increased proportion of lymphocytes and of LUCs.

five-part differential count including LUCs, reticulocyte count and haematocrit (HCT). Daily quality control (QC) on the instrument was performed using the high, low and normal 3-in-1 Advia TESTPointÔ controls (Siemens Healthcare Diagnostics). In addition to daily assessments, cumulative QC data were submitted to the Siemens Healthcare Diagnostics CHECKpointÔ Interlab Quality Control Program every 2 weeks for analysis.

Bone Marrow Microscopical Analysis

A board-certified veterinary clinical pathologist blinded to the results of the automated analysis performed 500-cell manual differential counts of stained BM smears. The following 15 categories of cells were enumerated: myeloblast + promyelocyte; neutrophilic myelocyte; neutrophilic metamyelocyte; band neutrophil; segmented neutrophil; eosinophilic

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

E. Tan et al.

4

granulocytes (all stages); basophilic granulocytes (all stages); rubriblast + prorubricyte; rubricyte; metarubricyte; lymphocyte; plasma cell; monocy te + monoblast; mast cell; and miscellaneous cells (Riley et al., 2009; Harvey, 2012). Lyzed cells were not enumerated, but were semiquantified as part of sample quality assessment. Granulocytic-toerythrocytic cell (GE) ratios were derived by dividing the number of granulocytic cells by the number of nucleated erythroid cells identified in the 500-cell differential count. Megakaryocytes were enumerated separately (total number in 10 random, nonoverlapping fields at 100 magnification). In addition to differential cell counting, smear cellularity, particle cellularity and particle number for each smear were scored by two clinical pathologists (Table 1). Where applicable, scoring of polychromatophil frequency and lymphocyte + plasma cell (LP) number was also performed. Each clinical pathologist scored smears independently. Statistical Analysis

Correlations between automated and manually determined variables were assessed by calculating Spearman’s rank correlation coefficients. Spearman’s rank correlation coefficients for automated WBC counts and smear cellularity, WBC counts and particle cellularity and WBC counts and particle number were calculated to determine whether

Table 1 Criteria for cytological assessment of bone marrow aspirate smears Bone marrow particles per slide* Score Particle cellularity† Score Smear cellularity‡ Polychromatophil frequencyx Granulocyte reserve{ Lymphocytes/plasma cells#

0

1e3

4e9

0 1 2 <25% 26e75% >75% 1 2 3 1 (poor) to 5 (excellent) 1 (low) to 5 (high)

>9 3

1 (low) to 5 (high) 1 (low) to 5 (high)

Adapted from Defarges et al., 2013. * 10 microscope objective. † Proportion of particles composed of haemopoietic cells versus fat. ‡ Approximate number of identifiable haemopoietic cells per smear: 1, <100; 2, 101e200; 3, 201e300; 4, 301e400; 5, >401. x Proportion of polychromatic RBCs (40 objective): 1, <1%; 2, 1e4%; 3, 5e9%; 4, 10e15%; 5, >15%. { Proportion of segmented, band and metamyelocyte neutrophils among all nucleated cells: 1, <5%; 2, 5e9%; 3, 10e20%; 4, 20e30%; 5, >30%. # Proportion of lymphocytes and plasma cells among all nucleated cells: 1, <2%; 2, 2e4%; 3, 5e9%; 4, 10e20%; 5, >20%.

Advia analysis could predict overall sample quality. Since there was no assumption regarding agreement between clinical pathologists, the average of the scores from the clinical pathologists was used in the correlation analysis. A Welch t-test was performed to determine whether there was a significant difference between automated WBC counts from samples assessed microscopically as ‘inadequate’ or ‘poorly cellular’ and those that were deemed ‘sufficiently cellular’ (i.e. smear cellularity score $3, Table 1). To assess the degree of concordance between clinical pathologists in qualitative assessments of BM smears, marrow granulocyte reserve (MGR; neutrophilic segmented cells + bands + metamyelocytes) and LP number were determined as outlined in Table 1. Concordance or agreement was assessed through calculation of kappa coefficients and since scores were ordinal in nature, weighted kappa was chosen (Altman, 1991). SAS/STAT software (version 9.2, SAS, Toronto, Ontario, Canada) was used to calculate weighted kappa scores and MedCalc (version 12.2.0, MedCalc Software, Mariekerke, Belgium) was used to perform the Welch t-test and to determine Spearman’s rank correlation and summary statistics. Clinical and Laboratory Findings and Outcome

After complete analysis of all BM samples, medical records of dogs included in the study were searched for the following information: initial presenting complaint(s), physical examination findings and haematological abnormalities at presentation, treatment (if any) and outcome. If final outcome could not be ascertained through the medical record, referring veterinarians were contacted by telephone and questioned about follow-up treatment, survival and, if deceased, date and cause of death. Advia Cytogram Analysis

The Advia 2120 generates graphical information for each of the WBC channel analyses in the form of cytograms, wherein cell clusters are easily visualized. We wished to determine whether visual comparison of cytograms from haematologically abnormal dogs with those from normal dogs would provide useful diagnostic information. After statistical analyses were complete, the samples were grouped according to diagnosis (myeloid neoplasia, metastatic/nonhaemopoietic neoplasia, lymphoid neoplasia, granulocytic hyperplasia, erythroid hyperplasia or BM hypoplasia/hypocellular sample) and the corresponding cytograms were compared with those of haematologically normal dogs.

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

5

Automated Bone Marrow Analysis in Dogs

leukaemia (ALL, n ¼ 4); chronic lymphocytic leukaemia (CLL, n ¼ 4); multiple myeloma (MM, n ¼ 3); myelodysplastic syndrome (MDS, n ¼ 1); chronic neutrophilic leukaemia (CNL, n ¼ 1); nonregenerative immune-mediated haemolytic anaemia (NRIMHA, n ¼ 5); immune-mediated neutropenia (IMNP, n ¼ 1); immune-mediated thrombocytopenia (IMTP, n ¼ 3); inflammatory disease (INFL, n ¼ 6); drug-induced myelotoxicity (MT, n ¼ 3) and normal haemopoiesis (n ¼ 2). AML was diagnosed when there was persistent uni- or multilineage cytopenia, infectious agents were not identified and BM blast cell count was $20% (Vardiman et al., 2009; Juopperi et al., 2011; Fig. 2A). The diagnosis of ALL was based on presence of persistent cytopenia, $20% BM blast cells, absence of solid lymphoid tumours and detection of one or several of the lymphoid markers CD3, CD4, CD5, CD8 or CD21 on flow cytometry or immunohistochemistry (IHC; Fig 2B). CLL was diagnosed by presence of lymphocytosis, lack of cytopenia or mild anaemia and flow cytometric detection of lymphocyte antigens. Cases of MM met at least two of the following four criteria: monoclonal gammopathy on serum protein electrophoresis, radiographical evidence of osteolysis, >10% plasma cells in BM and Bence-Jones proteinuria (Appel et al., 2008; Fig. 2C). The diagnosis of MDS was based on persistent cytopenia, dysplasia of blood and/or BM cells,

Results Cases

Samples from 51 dogs were available. In four cases the BM aspirate was clotted and in another four samples particles were not identified on smears. These samples were excluded, leaving 43 samples for full analysis. Samples originated from 28 (65%) males and 15 (35%) females, ranging in age from 11 months to 12 years (mean  SD; 6.9  2.9 years). Breeds included eight Labrador retrievers, seven mixed breed dogs, five golden retrievers, two rottweilers, two German shepherd dogs, two boxers and two Portuguese water dogs. One dog from each of the following breeds was also included: German short-haired pointer, cocker spaniel, mountain cur, Bernese mountain dog, Dogue de Bourdeaux, Cane Corso, Russian black terrier, English bulldog, Cairn terrier, border collie, whippet, dachshund, Staffordshire terrier, Great Pyrenees and soft-coated wheaten terrier. Diagnoses

In 23 cases the diagnosis from results of complete blood count (CBC) and/or BM smear review was neoplasia and in 20 cases it was non-neoplastic disease (Table 2). In addition to cytological BM preparations, sections of trephine biopsy samples were assessed in 21 cases. Diagnoses were acute myeloid leukaemia (AML, n ¼ 10); acute lymphocytic

Table 2 Patient characteristics and clinical findings in 43 dogs with haematological disorders Diagnosis: Number of cases Sex Male Female Age (years) Mean Range Presenting complaint Lethargy Anorexia Weight loss Vomiting, diarrhoea Bleeding diatheses Dyspnoea or cough Mass Other† Physical abnormalities Lymphadenopathy Organomegaly Fever Pallor

AML

ALL

CLL

CNL

MM

MDS

NRIMHA

IMTP

IMNP

INFLL

MT

NEO*

Total

10

4

4

1

3

1

5

3

1

6

3

2

43

6 4

4 0

0 4

1 0

2 1

1 0

4 1

2 1

1 0

4 2

3 0

0 2

28 15

6.3 2e12

8.6 2e11

9.5 7e11

10 e

8 6e10

3 e

7 3e11

5.3 5e6

9 e

5.7 4e8

2.1 1e4

7.5 7e8

6.9 1e12

10 7 1 2 0 0 0 1

4 3 1 1 0 1 0 1

1 1 1 0 0 1 0 0

1 1 0 0 0 0 0 1

1 1 0 0 1 0 0 1

1 1 0 1 0 0 0 0

5 4 1 0 0 0 0 0

3 2 0 0 3 0 0 0

1 1 1 1 0 0 0 1

6 4 0 2 0 1 0 1

2 3 0 2 2 0 0 0

1 1 0 1 0 0 1 0

36 29 5 10 6 3 1 6

0 1 4 0

1 0 1 0

0 1 0 0

0 0 0 0

0 0 0 0

0 0 1 0

1 2 0 5

1 1 0 3

1 0 1 0

2 0 6 0

1 0 2 0

0 0 0 0

7 6 15 8

*

BM sample obtained for tumour staging. Includes stiff gait and shifting lameness (AML and INFL), swollen limbs (ALL), ataxia (CNL), polyuria/polydipsia (MM), fever (IMNP).

†

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

E. Tan et al.

6

Fig. 2. Bone marrow smears from dogs with haematological disease. (A) AML. Note the high proportion of undifferentiated large round cells, smaller proportion of neutrophils, scattered rubricytes and relative paucity of polychromatic cells. 600. (B) ALL. A dense smear consisting of large round cells with rare plasma cells (arrowhead) and frequent mitotic figures (arrow). 400. (C) Multiple myeloma. Numerous plasma cells with moderate anisocytosis and anisokaryosis are scattered among granulocytes and rubricytes. There is one mitotic plasma cell (arrow). 600. (D) NRIMHA. There is an increased proportion of rubriblasts and prorubricytes and a paucity of metarubriyctes and polychromatophilic cells. 600.

NRIMA and reticulocytes ranged from 8 to 60  109/l. Dogs diagnosed with IMTP had marked thrombocytopenia, negative tick-borne disease serology and no underlying disease. BM smears from dogs with IMTP typically showed megakaryocytic and erythrocytic hyperplasia. IMNP was diagnosed when at least three of four of the following criteria were fulfilled (Perkins et al., 2004): (1) absence of other known causes of neutropenia; (2) presence of concurrent immune-mediated disease; (3) BM smears with granulocytic maturation arrest and/or left shift; and (4) improvement of neutropenia with immunosuppressive therapy. Diagnosis of infectious or INFL was based on relevant history, clinical signs and

hypercellular BM and BM blast count between 3 and 20%. The single dog with CNL had marked neutrophilia (>90  109/l) over a 1-month period, a moderate number of immature neutrophils in blood, no evidence of infectious or INFL and the neutrophilia was unresponsive to broad-spectrum antimicrobial therapy. BM smears showed marked granulocytic hyperplasia. NRIMA was diagnosed in dogs that had nonregenerative anaemia for >5 days, absence of other causes of anaemia, BM erythroid maturation arrest and/or hyperplasia and/or myelofibrosis (Stokol et al., 2000; Weiss, 2008; Fig. 2D). HCT ranged from 0.06 to 0.28 l/l (mean 0.14 l/l) in dogs with

Table 3 CBC results in dogs with diagnosis of leukaemia, myeloma or MDS Diagnosis

AML ALL CLL CNL MM MDS

WBC (109/l)

Neutrophils (109/l)

Lymphocytes (109/l)

Haematocrit (l/l)

Platelets (109/l)

Mean

Range

Mean

Range

Mean

Range

Mean

Range

Mean

Range

11.5 66.8 49.6 106.1 4.9 4.2

1.0e30.7 1.0e228.8 1.4e173.5 e 4.7e5.1 e

5.5 2.1 3.5 60.5 3.1 0

0.2e21.0 0.3e5.1 0.2e7.4 e 3.0e3.2 e

4.5 63.8 45.4 1.0 1.3 3.6

0.3e23.4 0.3e224.0 1.1e167.0 e 1.2e1.3 e

0.27 0.26 0.32 0.26 0.39 0.30

0.17e0.40 0.19e0.30 0.20e0.44 e 0.38e0.40 e

123 70 107 129 158 81

26e360 11e196 11e196 e 147e168 e

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

7

Automated Bone Marrow Analysis in Dogs Table 4 Correlation between automated and manual BM differential counts Automated analysis

Manual analysis

Proportion

Mean  SD

Proportion

Mean  SD

Neutrophils

30.7  24.0

Lymphocytes

41.7  20.6

Segmented cells Segmented cells/bands Segmented cells/bands/metamyelocytes Lymphocytes Lymphocytes/plasma cells Lymphocytes/plasma cells/rubricytes Lymphocytes/metarubricytes/rubricytes Monocytes Myeloblasts/promyelocytes Eosinophils Blast cells Myeloblasts/promyelocytes Rubriblasts/prorubricytes Polychromatophil score

10.0  8.7 20.8  18.5 23.8  22.7 27.0  14.7 18.6  35.3 38.1  52.3 41.5  25.7 1.2  1.2 8.9  4.9 2.2  2.1 7.0  4.1 8.9  4.9 4.1  3.2 2.4  1.3

Monocytes

6.8  4.9

Eosinophils LUCs*

3.1  2.6 11.4  9.3

Reticulocytes

62.2  53.9

Correlation coefficient (r)

P value

0.54 0.67 0.72 0.16 0.23 0.30 0.37 0.18 0.012 0.61 0.25 0.54 0.18 0.34

0.0005 <0.0001 <0.0001 0.3257 0.1695 0.0689 0.0177 0.2793 0.9420 0.0001 0.1270 0.0005 0.2845 0.0547

Bolded values indicate statistical significance. * Large unstained cells.

results of imaging, laboratory tests, exploratory surgery and culture and sensitivity of blood and/or body fluids. All dogs ultimately diagnosed with haematological malignancies presented with cytopenia of at least one cell line, most commonly anaemia (19 of 23 dogs had anaemia). WBC and differential counts were variable, with the highest count (228.8  109/l, reference limits 4.9e15.4  109/l) in a dog with ALL and the lowest counts (1.0  109/l) in two dogs with AML. The latter two dogs were considered ‘aleukaemic’, as atypical cells were not identified on blood smear review. Overall, dogs with ALL had higher mean WBC counts and lower mean platelet counts than dogs with AML (Table 3). Correlation between Automated and Manual Differential Counts

Spearman’s rank correlation coefficients (r) for automated and manual differential counts were calculated (Table 4). At P #0.05, automated percent neutrophils correlated with manual percent neutrophilic Table 5 Correlation between automated and microscopical WBC assessment Advia parameter

WBC count WBC count WBC count *

Microscopical parameter*

Correlation coefficient (r)

P value

Overall cellularity Particle cellularity Particle number

0.67 0.39 0.37

<0.0001 0.01 0.02

As defined in Table 1.

segmented cells (r ¼ 0.54), neutrophilic segmented cells + bands (r ¼ 0.67) and neutrophilic segmented cells + bands + metamyelocytes (MGR, r ¼ 0.72), with the strongest correlation between automated percent neutrophils and manual determination of MGR. Automated differential counting of ‘eosinophils’ also correlated with manual counts of eosinophilic granulocytes (r ¼ 0.61). Automated percent LUCs correlated with manual differential count of percent myeloblast + promyelocyte (granulocytic precursors, r ¼ 0.54), and automated percent lymphocytes correlated with manual percent of lymphocytes + rubricyte + metarubricytes (r ¼ 0.37). The later correlation was expected, as rubricytes and metarubricytes are peroxidase negative and may be similar in size to lymphocytes and thus would be considered as ‘lymphocytes’ by the Advia. There were no significant correlations between other analysed parameters. High standard deviation for some manual parameters such as lymphocytes/plasma cells resulted from their low frequency in all benign BM samples, but extremely high frequency in samples from dogs with ALL, CLL and some MM. The average manual score of the two clinical pathologists for each BM smear assessment was compared with the corresponding automated WBC count. Significant correlation was found between WBC count and smear cellularity (r ¼ 0.67, P <0.0001), particle cellularity (r ¼ 0.39, P ¼ 0.01) and particle number (r ¼ 0.37, P ¼ 0.02) (Table 5). Since WBC count correlated with smear cellularity, we wished to ascertain whether there was a significant difference between the WBC count of poorly cellular samples (smear cellularity score 1 or 2, n ¼ 4) and the WBC count of

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

E. Tan et al.

8 Table 6 Concordance between clinical pathologists on microscopical assessment of BM smears Parameter* Overall cellularity Particle cellularity Particle number Polychromasia Lymphocyte/plasma cell

Weighted kappa

Strength of agreement†

0.46 0.31 0.58 0.46 0.21

Moderate Fair Moderate Moderate Fair

*

As defined in Table 1. Based on Altman, 1991.

†

adequately cellular samples (smear cellularity score $3, n ¼ 39). Mean WBC  SD of hypocellular samples was 9.28  3.75, while that of adequately cellular samples was 133.58  115.85. Welch t-test showed a significant difference between WBC count of poorly cellular versus those of adequately cellular samples (t ¼ 6.834, DF ¼ 40.8, P <0.0001). Agreement between Clinical Pathologists

Microscopic assessment of smears was performed with the clinical pathologist blinded to the interpretations made by the other clinical pathologist. There was moderate agreement for cellularity, particle number and polychromasia, and fair agreement for particle cellularity and for LP (Table 6; Altman, 1991). Advia Cytograms

Advia analysis generated one basophil/lobularity cytogram and one peroxidase channel cytogram for each BM sample. Cytograms were grouped according to diagnosis. Since WBC counts in most BM samples were much higher (mean  SD ¼ 130.5  116.2  109/l) than WBC counts in peripheral blood (Fig. 1A), most Advia cytograms were very dense (Figs. 1BeH). Cellularity of samples was assessed readily on visual inspection of cytograms, which indicated relatively hypocellular (Fig. 1B) or highly cellular samples (Figs. 1CeG). Accordingly, there were significant correlations between WBC count and sample cellularity and particle cellularity and particle number. Cytograms of samples with granulocytic hyperplasia (Fig. 1D) were similar to those of normal (Fig. 1A) and CNL BM samples (Fig. 1E), except that there was a denser LUC region in CNL. Cytograms from ALL (Fig. 1F) and AML (Fig. 1G) samples had very dense lysis-resistant and LUC cell clusters, but a relative lack of cells with peroxidase content suggested ALL rather than AML. Cytograms from samples with CLL (Fig. 1H) were similar to cellular samples from normal dogs (Fig. 1C) except

for greater cell density in the lymphocyte and LUC region. Outcome

Most dogs diagnosed with acute leukaemia had short survival times, regardless of treatment (Table 7). The majority of dogs diagnosed with AML died or were humanely destroyed due to clinical deterioration within 6 months of diagnosis. Two dogs survived $2 years post diagnosis. One developed kidney failure of unknown aetiology and was humanely destroyed due to progressive disease approximately 25 months after diagnosis; another dog was alive and haematologically normal after 3 years. Three of four dogs diagnosed with ALL were humanely destroyed due to poor clinical condition within 2 months of diagnosis; one was discharged and lost to follow-up. Outcomes for dogs with CLL were more variable than for dogs with ALL. CLL was an incidental finding in three of four dogs. Consistent with previous reports (Vernau and Moore, 1999; Tasca et al., 2009; Comazzi et al., 2011; Valli et al., 2013), dogs with CLL had long survival times or died from unrelated causes. In one case (CD8+ T-CLL), lymphocytosis was an incidental and persistent (w2 yrs) finding and extensive clinical investigation ruled out other causes. Another case (CD8+ T-CLL) presented for respiratory distress secondary to pulmonary metastases of a previously diagnosed mammary carcinoma and a CBC disclosed lymphocytosis. A third case (B-CLL) presented for geriatric dentistry; lymphocytosis and hyperglobulinaemia were incidental findings on pre-anaesthetic blood work. Finally, a single dog with large granular lymphocytic CLL presented with lethargy and inappetence. Moderate to severe pancytopenia was found on CBC and despite treatment, the dog died at home 2 weeks after initial diagnosis. A necropsy examination was not performed.

Discussion Results from this and a prior study (Tan et al., 2013) indicate that automated BM analysis can provide valuable ‘pre-microscopical’ information, including estimates of MGR and rubricyte stores. A low WBC count on automated analysis corresponded with poorly cellular BM smears, while a high WBC count corresponded with adequately cellular and therefore diagnostically useful samples. The rapid turnaround time of automated analysis could inform the clinician of potentially poorly cellular or non-diagnostic samples and additional attempts at sample procurement could then be made prior to anaesthetic recovery

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

Duration of illness (months)

Breed

Age (years)

Haematological abnormalities

Therapy

Time to humane destruction (months)

AML 1 0.5 0.75 0.75 1 0.1 1 2 0.2 1

Labrador retriever German shepherd dog German short-haired pointer German shepherd dog Portuguese water dog Bulldog Great Pyrenees Labrador retriever Labrador retriever Golden retriever

12 7 4 7 9 4 4 7 2 9

Anaemia, neutropenia Pancytopenia Pancytopenia Pancytopenia Anaemia, neutropenia Pancytopenia Anaemia, thrombocytopenia Pancytopenia Pancytopenia Pancytopenia

Antimicrobials, glucocorticoids MAC† Antimicrobials, MAC Antimicrobials, L-asparaginase Antimicrobials, glucocorticoids None Antimicrobials MAC Antimicrobials, L-asparaginase Glucocorticoids

25* 0.75 0.5 0.1 0.1 0 NA‡ 6 >36 0.5

ALL 0.5 NA

Mixed breed Labrador retriever

11 10

Glucocorticoids Glucocorticoids

NA 1

2 NA

Golden retriever Rottweiler

10 10

Pancytopenia Anaemia, lymphocytosis, thrombocytopenia Pancytopenia Anaemia, lymphocytosis

Antimicrobials, MAC MAC

2 1

T-CLL 0.5 24 0.75

Cocker spaniel Golden retriever Golden retriever

10 11 7

Anaemia, lymphocytosis Anaemia, lymphocytosis Pancytopenia

Antimicrobials Glucocorticoids, L-asparaginase Glucocorticoids

8x 1 0.5

B-CLL NA

Mixed breed

10

Neutropenia, thrombocytopenia

Glucocorticoids, chlorambucil

>67

CNL 1

Labrador retriever

10

Anaemia, neutrophilia

Antimicrobials

NA

IMNP 2

Labrador retriever

9

Anaemia, neutropenia

Glucocorticoids, azathioprine

24{

INFL 3 3 0.2 3 0.2 0.75

Rottweiler Bernese Mountain dog Mixed breed Boxer Russian terrier Portuguese water dog

8 5 5 4 8 4

Anaemia, neutrophilia Anaemia, neutrophilia, left shift Neutropenia, thrombocytopenia Thrombocytopenia Neutropenia Pancytopenia

Antimicrobials Glucocorticoids, ciclosporin Antimicrobials Glucocorticoids, lymphadenectomy Antimicrobials Antimicrobials, transfusion

Unknown >12 NA NA NA NA

IMTP 0.5 0.2 0.2

Cairn terrier Border collie Mixed breed

5 6 5

Monocytosis, anaemia, thrombocytopenia Thrombocytopenia, Anaemia, neutrophilia Anaemia, thrombocytopenia

Glucocorticoids, ciclosporin, chlorambucil Glucocorticoids, ciclosporin Glucocorticoids, ciclosporin

<1 NA 2

9

(Continued on next page)

Automated Bone Marrow Analysis in Dogs

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

Table 7 Outcome for 43 dogs with haematological disorders

10

Duration of illness (months)

Breed

Age (years)

Haematological abnormalities

Therapy

Time to humane destruction (months)

MM 0.2 2

Staffordshire terrier Mixed breed

10 6

None None

10 2

0.75

Labrador

8

NA

Glucocorticoids, melphalan Glucocorticoids, melphalan, cyclophosphamide Glucocorticoids

Wheaten terrier

3

Anaemia

8‡

0.2 0.5

Mixed breed Mountain cur

6 11

Pancytopenia Anaemia

0.75

Dachshund

4

Anaemia, thrombocytosis

Whippet

7

Anaemia, neutrophilia

Prednisone, ciclosporin, packed RBCs, doxycycline Azathioprine, antimicrobials, transfusion Antimicrobials, immunosuppressive drugs, transfusion Antimicrobials, immunosuppressive drugs, transfusion Antimicrobials, immunosuppressive drugs, transfusion

NEO 0.5 None

Dogue de Bordeaux Labrador retriever

5 8

Thrombocytopenia No abnormalities

Multiagent chemotherapy Surgery, radiation, analgesia

1‡ 3‡

MT 0.1 0.2 1

Mixed breed Golden retriever Boxer

1 2 4

Neutropenia, thrombocytopenia Neutropenia Neutropenia, thrombocytopenia

Antimicrobials Antimicrobials Glucocorticoids, antimicrobials

NA NA NA

Cane Corso

3

Pancytopenia

Antimicrobials

2

NRIMHA 2

3

MDS 1 *

Humanely destroyed due to causes unrelated to leukaemia. Multiagent chemotherapy. ‡ Not applicable; dog still alive. x Died at home. { Humanely destroyed due to causes unrelated to IMNP. †

2

NA NA NA NA

E. Tan et al.

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

Table 7 (continued )

Automated Bone Marrow Analysis in Dogs

11

Fig. 3. (A) Peroxidase activity and nuclear density analysis (PANDA) to categorize cells by size and peroxidase activity. P0, large cells with no peroxidase activity (LUCs); P1, large cells with low peroxidase activity; P2, large cells with low to moderate peroxidase activity; P3, medium size cells with low to moderate peroxidase activity; P4, variably sized cells with moderate to high peroxidase activity; P5, large cells with high peroxidase activity; P6, medium size cells with high peroxidase activity (adapted from Gibbs, 2011). (B) Peroxidase cytogram of BM sample from a dog with AML showing P5 pattern indicated by a circle.

of the patient. A receiver-operating characteristic (ROC) curve analysis to establish cut-off values for WBC count might have been helpful; however, since only a small number of ‘poorly cellular’ samples (n ¼ 4) were included in this study, such analysis would not be meaningful (Greiner et al., 2000; Obuchowsky et al., 2004). Nonetheless, BM WBC counts above those of normal peripheral blood (>11.2  109/l) appear important to yield a microscopically useful sample, unless blast cells predominate. Clots or lack of particles in BM aspirates were reasons for eliminating samples from analysis. Sample clotting in four samples was most likely related to operator experience and factors such as size and body score of patients (Abrams-Ogg et al., 2013). Notwithstanding operator experience, inadequacy of samples was also reported for 3.9%e7.7% of BM aspirates from human patients and was most commonly ascribed to BM fibrosis, hypercellularity or both, as diagnosed from trephine biopsy (Engeset et al., 1979; Navone and Colombano, 1984; Humphries, 1990). Possible reasons for the lack of BM particles in aspirates may be similar to those in people. Three dogs lacked particles in aspirates and two of these had hypercellular BM, while one had MF diagnosed on histopathology review of trephine biopsy sections. Review of Advia cytograms revealed differences between ‘normal’ BM samples and neoplastic samples and distinct and repeatable patterns amongst diagnostic groupings. Our findings corroborate those of Bauer and Moritz (2002) using the Advia 120, but herein we also included samples from haematologically normal dogs as well as those with non-

neoplastic conditions. The results of the present study indicate that visual inspection of Advia cytograms of BM may flag samples with acute leukaemia. Alterations in Advia cytogram patterns in peripheral blood from leukaemic human patients have been described extensively in a system known as peroxidase activity and nuclear density analysis (PANDA) (D’Onofrio et al., 1987; Gibbs, 2011). Although originally intended for analysis of peripheral blood, PANDA has potential for analysis of BM fluid. In particular, the PA of leucocytes may be divided into seven approximate categories, from P0 to P6 (Fig. 3A), which correspond with the level of leukaemic cell differentiation. The P0 cluster indicates large cells with no PA (equivalent to a LUC cluster paralleling the y-axis), P5 indicates very large cells with very high PA and P6 indicates cells of intermediate size with very high PA. In addition to dense clusters in other regions, cells in P0 were present in all BM samples of adequate cellularity, due to the presence of large blast cells of various lineages lacking PA. Most cytograms of AML samples had P5 predominance with cell cluster extension into P4 and P3 (Fig. 3B). These findings are consistent with a large number of myeloblasts, promyelocytes and myelocytes having a continuum of PA. In contrast, typical ALL cytograms (Fig. 2F) had very dense P0 patterns with few cells in other regions. Not surprisingly, for cases with a high proportion of Advia LUCs, such as all cases of ALL (LUC $60%), manual enumeration also yielded a high proportion ($65%) of undifferentiated blast cells (data not shown). Lack of distinction between small lymphocytes and rubricytes by automated analysis necessitates their distinction on

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

12

E. Tan et al.

microscopical slide review and also emphasizes that blood reticulocyte counts form the basis for decisions regarding regeneration of anaemia. Chronic leukaemia did not consistently have recognizable cytogram alterations. In the PANDA system, chronic myelogenous leukaemia (CML) in human peripheral blood was described as having a more pronounced P3 or P4 pattern (D’Onofrio et al., 1987). Similar changes were not apparent in cytograms from canine CNL, which was also indistinguishable from granulocytic hyperplasia. Cytograms of CLL were similar to those of normal BM, except for increased density in the small size and no PA cell cluster. Over 50% of samples were from dogs diagnosed with haematological neoplasia and 43% were from dogs with AML. This high proportion likely reflects the need for BM analysis to diagnose AML rather than disease prevalence. Most dogs with AML had short survival times after diagnosis, but two dogs survived for 2 years despite only brief chemotherapy. This finding may indicate that the diagnostic criteria applied were insufficiently stringent; that these dogs had temporary, severely altered but benign haemopoiesis; that conditions such as transient myeloproliferative disorder of people may also occurs in dogs (Gamis and Smith, 2012); or that AML may not have an invariably grave prognosis. Survival of similar duration was reported in a dog with acute megakaryoblastic leukaemia treated extensively with chemotherapy (Willmann et al., 2009) and suggests that AML in dogs is a more heterogeneous tumour than often appreciated (Juopperi et al., 2011). Agreement regarding the final interpretation of BM smears was high between clinical pathologists, but concordance on qualitative assessment of certain components of the slides was moderate. This may reflect the inherent subjectivity of microscopical BM evaluation (Travlos, 2006) and might be improved if more objective assessment of ‘sample quality’ from automated analysis is also considered. The limitations of this study pertain to relatively small sample size and the lack of uniformity of treatment for neoplasia and other haematological diseases. However, the results demonstrate that analysis of canine BM samples in the Advia 2120 provides useful and expedient ‘pre-microscopical’ screening in haematologically abnormal dogs. The WBC count yielded valuable information about diagnostic sample quality. Advia differential counts provided information about the adequacy of granulocytic and rubricytic stores and visual inspection of Advia cytograms identified samples with likely acute leukaemia. The cell viability of anticoagulated BM samples is probably as limited as that of blood

samples, and samples would require similar handling for automated analysis. Finally, automated methods cannot supplant microscopical examination of BM.

Acknowledgements The authors thank W. Sears, Department of Population Medicine, University of Guelph, for valuable expertise and assistance in statistical analysis; N. Lemieux for her help with the figures; and A. Prieto, Animal Health Laboratory, University of Guelph, for technical support with the Advia 2120.

References Abrams-Ogg A, Defarges A, Foster R, Bienzle D (2013) Comparison of canine core biopsies from multiple sites using different techniques and needles. Veterinary Clinical Pathology, 41, 235e242. Altman G (1991) Practical Statistics for Medical Research. Chapman & Hall, London, pp. 396e409. Appel SL, Moens NM, Abrams-Ogg A, Woods JP, Nykamp S et al. (2008) Multiple myeloma with central nervous system involvement in a cat. Journal of the American Veterinary Medical Association, 233, 743e747. Bauer N, Moritz A (2002) Durchflusszytometrische Knochenmarkuntersurchung mit dem H€ amatologiesystem ADVIA 120. Tier€aztliche Praxis, 30, 472e476. Bienzle D (2012) Collection and interpretation of bone marrow samples. In: Manual of Canine and Feline Haematology and Transfusion Medicine, 2nd Edit., MJ Day, B Kohn, Eds., British Small Animal Veterinary Association, Quedgeley, pp. 21e30. Comazzi S, Gelain ME, Martini V, Riondato F, Miniscalco B et al. (2011) Immunophenotype predicts survival time in dogs with chronic lymphocytic leukemia. Journal of Veterinary Internal Medicine, 25, 100e106. Defarges A, Abrams-Ogg A, Foster R, Bienzle D (2013) Comparison of sternal, iliac, and humeral bone marrow aspiration in small dogs. Veterinary Clinical Pathology, 42, 170e176. D’Onofrio G, Mancini S, Leone G, Bizzi B, Mango G (1987) Identification of blast cells in peripheral blood through automatic assessment of nuclear density: a new tool for monitoring patients with acute leukaemia. British Journal of Haematology, 66, 473e477. Engeset A, Nesheim A, Sokilowski J (1979) Incidence of ‘dry tap’ on bone marrow aspirations in lymphomas and carcinomas: diagnostic value of the small material in the needle. Scandinavian Journal of Haematology, 22, 417e422. Fan G, Alvares C, Ismail S, Perrotta G, Robert L et al. (1999) Quantitative and qualitative bone marrow analysis using the Abbott Cell-Dyn 4000 hematology analyzer. Laboratory Hematology, 5, 45e51. Gamis A, Smith F (2012) Transient myeloproliferative disorder in children with Down syndrome: clarity to this enigmatic disorder. British Journal of Haematology, 159, 277e287.

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005

Automated Bone Marrow Analysis in Dogs

Gibbs G (2011) White cells. In: Advia 120/2120i: a Guide to Cytogram Interpretation, Rossendale Books, Lancashire, pp. 5e31. Greiner M, Pfeiffer D, Smith RD (2000) Principles and practical application of the receiver-operating characteristic analysis for diagnostic tests. Preventive Veterinary Medicine, 45, 23e41. Harvey J (2012) Bone marrow examination. Veterinary Hematology: A Diagnostic Guide and Color Atlas, Elsevier Saunders, St Louis, pp. 234e259. Humphries JE (1990) Dry tap bone marrow aspiration: clinical significance. American Journal of Hematology, 35, 247e250. Juopperi T, Bienzle D, Bemreuter D, Vernau W, Thrall MA et al. (2011) Prognostic markers for myeloid neoplasms: a comparative review of the literature and goals for future investigation. Veterinary Pathology, 48, 182e197. Lillieh€o€ok I, Tvedten H (2011) Errors in basophil enumeration with 3 veterinary hematology systems and observations on occurrence of basophils in dogs. Veterinary Clinical Pathology, 40, 450e458. Navone R, Colombano MT (1984) Histopathological trephine biopsy findings in cases of ‘dry tap’ bone marrow aspirations. Applied Pathology, 2, 264e271. Obuchowsky NA, Lieber ML, Wians FH (2004) ROC curves in clinical chemistry: uses, misuses and possible solutions. Clinical Chemistry, 50, 1118e1125. Perkins MC, Canfield P, Churcher RK, Malik R (2004) Immune-mediated neutropenia suspected in five dogs. Australian Veterinary Journal, 82, 52e57. Provencher-Bolliger A (2004) Cytologic evaluation of bone marrow in rats: indications, methods, and normal morphology. Veterinary Clinical Pathology, 33, 58e67. Riley R, Williams D, Ross M, Zhao S, Chesney A et al. (2009) Bone marrow aspirate and biopsy: a pathologist’s perspective. II. Interpretation of the bone marrow aspirate and biopsy. Journal of Clinical Laboratory Analysis, 23, 259e307. Stokol T, Blue JT, French TW (2000) Idiopathic pure red cell aplasia and non-regenerative immune-mediated

13

hemolytic anemia in dogs: 43 cases. Journal of the American Veterinary Medical Association, 216, 1429e1436. Tan E, Abrams-Ogg A, Defarges A, Bienzle D (2013) Automated hematological analysis of bone marrow aspirate samples from healthy beagle dogs. Veterinary Clinical Pathology. in press. Tasca S, Carli E, Caldin M, Menegazzo L, Furlanello T et al. (2009) Hematologic abnormalities and flow cytometric immunophenotyping results in dogs with hematopoietic neoplasia: 210 cases (2002e2006). Veterinary Clinical Pathology, 38, 2e12. Travlos GS (2006) Normal structure, function and histology of the bone marrow. Toxicologic Pathology, 34, 248e262. Valli VE, Kass PH, Myint MS, Scott F (2013) Canine lymphomas: association of classification type, disease stage, tumor subtype, mitotic rate, and treatment with survival. Veterinary Pathology, 50, 738e748. Vardiman JW, J€ uergen T, Arber DA, Brunning RD, Borowitz MJ et al. (2009) The 2008 revision of the World Health Organization (WHO) classification of myeloid neoplasms and acute leukaemia: rationale and important changes. Blood, 114, 937e951. Vernau W, Moore PF (1999) An immunophenotypic study of canine leukaemias and preliminary assessment of clonality by polymerase chain reaction. Veterinary Immunology and Immunopathology, 69, 145e164. Weiss DJ (2008) Bone marrow pathology in dogs and cats with non-regenerative immune mediated hemolytic anemia and pure red cell aplasia. Journal of Comparative Pathology, 138, 46e53. ullauer L, Schwendenwein I, Willmann M, M€ Wolfesberger B, Leiter M et al. (2009) Chemotherapy in canine acute megakaryoblastic leukaemia: a case report and review of the literature. In Vivo, 23, 911e918.

October 16th, 2013 ½ Received, Accepted, February 14th, 2014 

Please cite this article in press as: Tan E, et al., Automated Analysis of Bone Marrow Aspirates from Dogs with Haematological Disorders, Journal of Comparative Pathology (2014), http://dx.doi.org/10.1016/j.jcpa.2014.02.005