Optimizing antibody panels for efficient and cost-effective flow cytometric diagnosis of acute leukemia

Authors


  • How to cite this article: Haycocks NG, Lawrence L, Cain JW, Zhao XF. Optimizing antibody panels for efficient and cost-effective flow cytometric diagnosis of acute leukemia. Cytometry Part B 2011;80B: 221–229.

Abstract

Background:

Differentiating acute myeloid leukemia (AML) from acute lymphoblastic leukemia (ALL) determines effective patient management and often depends on flow cytometry. Antibodies used in flow cytometry are costly, and the expenses are not always reimbursed. Having observed that AML and ALL have distinct patterns in the CD45/SSC panel, we set to analyze more leukemia cases and establish an algorithm for the efficient diagnosis of acute leukemia.

Methods:

We retrospectively analyzed 127 consecutive cases of acute leukemia within the last 2 years and correlated the blast distribution patterns in the CD45/SSC panel, with the morphology and the detailed immunophenotype.

Results:

Our results show that all the acute leukemias can be initially triaged into AML, ALL, and Indeterminate provisional groups based on the blast distribution patterns in the CD45/SSC panel and morphology. Each group was then further analyzed with tailored AML, ALL, and Indeterminate flow panels. Using this approach, we have efficiently and correctly diagnosed almost all the acute leukemias. Our analysis also determined the minimal numbers of immunological markers needed for the lineage assignment of acute leukemia.

Conclusion:

The algorithmic approach with tailored subsequent antibody selection could maintain diagnostic accuracy while significantly reducing reagent use, labor, and time. With a shrinking reimbursement for flow cytometric studies, an increase in laboratory efficiency without compromising diagnostic accuracy or turnaround time will contribute to preserving revenue and optimizing clinical service. © 2011 International Clinical Cytometry Society

Flow cytometry is an indispensible tool for the diagnosis of acute leukemia. With increasing numbers of high-quality monoclonal antibodies that recognize the various hematopoietic cell markers becoming commercially available, routine flow cytometry panels are expanding in many diagnostic laboratories. Inherent to this trend is a concomitant increase in reagent cost and labor necessary to perform these panels. Despite this, clinical treatment relies predominantly on one important diagnostic distinction: the differentiation of acute myeloid leukemia (AML) from acute lymphoblastic leukemia (ALL).

Flow cytometry is a very complex laboratory procedure that requires both great technical skill and very knowledgeable and experienced interpretation. Because of the variations from laboratory to laboratory, it is extremely difficult to standardize the procedures and antibody panels. In the last decade, many flow cytometry laboratories in the world have endeavored to find the minimal number of antibodies that could correctly diagnose acute leukemia (1–3). In 2006, a group of international experts met in Bethesda to attempt to define a consensus set of reagents suitable for general use in the diagnosis and monitoring of hematopoietic neoplasms (1). Consensus reagents for initial evaluation for hematopoietic neoplasms include 7 B-cell markers, 8 T-cell markers, and 12 myelomonocytic markers. Fourteen to 23 marker panels were used for secondary evaluation of specific hematopoietic cell lineages.

In an era of ever increasing demand for testing capabilities with static or declining funding (4), daily practice often faces a dilemma of diagnosing acute leukemia accurately and quickly, while cutting the cost of reagents and labor. Because effectively triaging specimens to select more tailored antibody panels could theoretically reduce unnecessary reagent use without compromising each diagnosis, seeking an efficient means to perform this triage is necessary. Having observed reproducible variation between lymphoid and myeloid blasts, and distinct blast distribution patterns between AML and ALL when plotted by CD45 intensity versus side scatter (SSC), we set out to establish an efficient algorithm for distinguishing AML from ALL with minimal immunological markers.

MATERIALS AND METHODS

Patients and Specimens

With the University of Maryland School of Medicine Institutional Review Board approval, 127 consecutive cases of newly diagnosed acute leukemia were retrieved from the archives of The University of Maryland Medical Center (UMMC) Department of Pathology, dating from April 2008 to March 2010. The samples consisted of 79 peripheral blood and 48 bone marrow specimens. The patients included 61 males and 66 females (M:F ratio of 0.92), ranging from 1 to 87 years of age (median 52 years) (Table 1). Cases were included if they met the criteria of the 2008 WHO Classification for acute leukemia (5); this consisted of both de novo acute leukemias as well as those arising from preexisting myeloproliferative neoplasm and/or myelodysplastic syndrome. Cases were excluded if they represented relapse from a previously diagnosed acute leukemia. One case of acute leukemia was also excluded, because the aspirate was insufficient for morphologic examination. Another case of T-cell prolymphocytic leukemia was excluded as it is considered a mature T-cell neoplasm (5).

Table 1. Demographic Data of Patients with New Acute Leukemia
LeukemiasNMedian age (range)M:F
All12752 (1–87)0.92
AML10056 (6–87)0.85
 Monocytic2757 (15–86)1.1
 Other7356 (6–87)0.78
ALL2728 (1–78)1.3
 B-cell2222 (1–66)1.2
 T-cell542 (25–78)1.5

Routine and Special Stains

Peripheral blood smears were stained with Wright stain, whereas the marrow smears were stained with Wright–Giemsa stain (6). Most of the smears were also cytochemically analyzed by myeloperoxidase (MPO), Sudan black B, nonspecific esterase (α-naphthyl-butyrate), and periodic-acid-Schiff (PAS) stains (7). Marrow core biopsies were routinely stained with hematoxylin-eosin and PAS stains. Paraffin immunoperoxidase stains were performed using a Biotek Techmate 1000 autostainer (Ventana Medical Systems, Tucson, AZ) according to the manufacturer's protocol. Mouse anti-human CD3, CD5, CD20, CD79a, CD34, MPO, and TdT (DAKO Corporation, Carpinteria, CA) were used as the primary antibodies. Horseradish peroxidase-labeled rabbit anti-mouse polyclonal antibodies were used to convert the chromogen 3,3′-diaminobenzidine tetrahydrochloride. All the stainings were performed with appropriate positive and negative controls.

Flow Cytometry

Peripheral blood and bone marrow samples were obtained from each patient before treatment and processed according to the International Society for Advancement of Cytometry guideline (8) and a Standard Operating Procedure (SOP) of the UMMC Flow Cytometry Laboratory. Briefly, peripheral blood was processed directly with a single wash to wash off excess plasma, and fluorochrome-labeled monoclonal antibodies were directly added after this step. Bone marrow sample was processed with a selective lysis of red blood cells with 15–20-ml ACK lysis buffer (BioSource, International Biofluids Division, Rockville, MD). After washing with phosphate-buffered saline (PBS), the marrow cells were resuspended with 1 ml of PBS. For a full leukemia panel, 28 different monoclonal antibodies were pipetted in 13 different tubes. Each antibody in each tube was differentiated by one of four tagged fluorochromes: FITC, PE, PerCP, and APC (Table 2). The binding reaction was incubated at 20 min, followed by a 5-min lysing procedure with 0.75 ml of lysis buffer and one last wash step (9, 10). Nuclear and cytoplasmic stains were performed using a Fix and Perm Kit according to a standard protocol (Caltag Laboratories, Burlingame, CA).

Table 2. Antibodies Used in Flow Cytometry Analysis
AntibodiesClonesFluorchromesSources
  1. Footnote: FITC, fluorescein isothiocyanate; PE, phycoerythrin; PerCP, peridinin chlorophyll protein; APC, allophycocyanin.

CD2S5.2FITC, APC, PEBD
CD3SK7FITC, APC, PE, PerCPBD
CD4Leu-3aFITCBD-Simultest
CD5L17F12FITC, APCBD
CD7M-T701FITCBD
CD8Leu-2aAPC, PEBD
CD10W8E7FITCBD
CD11bD12PEBD
CD13L138PEBD
CD14M0P9PEBD
CD15MMAFITCBD
CD163G8AlexcFluor647BD
CD19SJ25C1FITC, APC, PE, PerCPBD
CD20L27FITC, APC, PE, PerCPBD
CD22S-HCL-1APCBD
CD23EVBVC5-5PEBD
CD33P67.6FITC, APC, PEBD
CD34My10FITC, APC, PEBD
CD38HB7APC, PEBD
CD45J2D1FITC, APC, PE, PerCPBD
CD56NCAM16.2FITC, PEBD
CD6410.1FITCBD
CD71L01.1FITCBD
CD117104D.2104D.2BD
cCD3SK7FITC, APC, PE, PerCPBD
cCD79aHM47PEBD
FMC7FMC7FITCBD
HLADRL243APCBD
KappaTB28-2FITCBD
Lambda1-155-2PEBD
MPO5B8FITCBD
TdTUnknownFITCSupertechs

After the last wash, the cell buttons in each tube were resuspended with 0.25 ml PBS and subjected to data acquisition and analysis. Four-color flow cytometric analysis was performed using a BD FACS Calibur instrument (Becton-Dickinson, Franklin Lakes, NJ) according to a standard protocol. Ten thousand events were acquired per tube to ensure the best definition of each cell population. The acquired data were analyzed using CellQuest software (Becton-Dickinson, San Jose, CA). Each tube was analyzed to determine which cell populations show positive, negative, or variances between the two. After regating and analysis was complete, a summing of the positive antibodies is conducted and entered into the final report as long as duplicate antibodies were in agreement. Whenever there were disagreements between the duplicate antibodies, the assays were repeated using fresh antibodies with the correct fluorochrome, which fixed the problem. Positivity was defined as staining on >20% of the cell population.

Case Review Criteria

Based on our observation that the blasts of AML are usually localized as a round/oval cluster and the blasts of ALL distribute longitudinally along the CD45 axis in the CD45/SSC flow cytometry panel (Fig. 1), two hematopathologists (NGH and XFZ) blind to the diagnoses of leukemia independently reviewed the dot plots in the CD45/SSC panel and rendered a presumptive diagnoses of AML, ALL, or Uncertain for each case. The peripheral blood or marrow aspirate smears were independently reviewed by the two hematopathologists (reviewers), and each was presumptively classified as AML, ALL, or Uncertain as well.

Figure 1.

Blast distribution patterns in the CD45/SSC panel. A: Schematic diagram of normal gates (upper left) for lymphoid (Ly), myeloid/granulocytic (My), and monocytic (Mo) lineages. Blasts (solid gray) from each lineage tend to cluster with distinct patterns in the CD45/SSC panel. B: Examples of blast distribution pattern for acute lymphoid (left), myeloid (right upper and right center), and monocytic (right lower) leukemias in the CD45/SSC panel.

The classification of each case based on morphology was performed according to standard descriptions of blast morphology (5). In general terms, blasts that showed relatively abundant cytoplasm, cytoplasmic granules, Auer rods, round to oval nuclei, fine chromatin with conspicuous nucleoli were considered myeloblasts. Blasts with abundant occasionally vacuolated grey cytoplasm, round nuclei with delicate lacy chromatin, and prominent nucleoli were considered monoblasts. Lymphoblasts were considered if the blasts had scant cytoplasm, round or irregular nuclei, slightly condensed chromatin, and inconspicuous nucleoli. In cases of leukemic cells with features of both myeloblasts and lymphoblasts, thus defined as Uncertain, the provisional diagnoses will rely solely on the “certain” CD45/SSC flow cytometric patterns.

When the morphologic and flow cytometric presumptive diagnoses agreed, the leukemia was assigned a provisional diagnosis of either AML or ALL. When the morphologic and flow cytometric presumptive diagnoses did not agree or review of both morphology and flow cytometry could not render a provisional diagnosis, the lineage was considered Indeterminate. When either CD45/SSC pattern or morphology was considered “insufficient” or “uncertain,” we assigned a provisional diagnosis based on the other “certain” information. If both CD45/SSC and morphology were “uncertain,” the leukemia was considered Indeterminate.

RESULTS

Distinct Patterns of AML and ALL in CD45/SSC Panel

We observed that blasts are usually moderate, dim, or negative for CD45, depending on the stage of differentiation and maturation. Distinct blast distribution patterns have been identified in the CD45/SSC panel for AML, ALL, and acute monocytic leukemia (Fig. 1A). With our instrument settings, the myeloblasts usually stay in the so-called blast gate—a roughly round or oval area with dim to moderate CD45 expression and low-side scatter. Myeloblasts often display small variation in CD45 expression and production of primary granules (11), thus usually generating a round or oval clustering pattern by CD45/SSC (Fig. 1B). When normal myeloblasts mature, they shift right in the CD45/SSC panel as they synthesize more granules. This rightward shift is recapitulated in myeloid leukemic cells with granules, such as acute promyelocytic leukemia and acute myeloblastic leukemia with maturation. When there is monocytic differentiation in the blasts, they often acquire slightly more CD45 expression than usual myeloblasts do; the neoplastic cells will cluster upwards and to the right, residing in the region between the myeloblasts and mature monocytes (Fig. 1B). Because residual maturating myeloid precursors may be present or leukemic myeloblasts may show some maturation, mono/myeloblasts tend to show close proximity to the normal monocytes/granulocytes and/or confluence with one or more of those populations. In erythroleukemia, although the eythroblasts also originate from the myeloid progenitors in the blast gate region, while maturing they shift straight downward with loss of CD45 expression. The erythroblast distribution sometimes can overlap with that of the lymphoblasts (see below). With the maturation of myeloblasts, stem-cell marker CD34 and HLA-DR are usually lost (12), and there is commensurately increased production of CD13, CD33, and MPO. However, when the blasts show monocytic differentiation, they lose CD34, but keep HLA-DR, and acquire monocytic markers including CD4, CD14, and CD64. When the blasts mature toward the erythroid lineage, they lose CD34 and CD45, but acquire glycophorin A and hemoglobin (13).

Because lymphoid cells usually lack cytoplasmic granules, lymphoblasts are more uniform on the SSC axis. However, lymphoblasts show larger variation in surface CD45 expression; thus, they often distribute longitudinally along the CD45 axis (Figs. 1A and 1B). Lymphoblasts typically have clear demarcation from maturing granulocytes, producing a slender, elongated distributing pattern that was associated with mature lymphocytes. Precursor B-cells express CD19, and while maturing, they gradually acquire CD20 and immunoglobulins on their surfaces. Thus, only a subset of B-ALL is CD20-positive and/or expresses surface immunoglobulin. T cells express CD3; negative surface CD3/positive cytoplasmic CD3 is consistent with T lymphoblasts, whereas CD4/CD8 double positivity helps exclude mature T-cell leukemia/lymphoma (except for T-cell prolymphocytic leukemia).

Despite these observations, a subset of cases showed conflicting features, making them difficult to be classified (thus defined Indeterminate).

Correlation Between Flow Cytometric Pattern and Morphology

The presumptive diagnosis by CD45/SSC pattern alone was correct in 94.1% of the cases, incorrect in 4.3%, and Uncertain in 1.6% (average of the two hematopathologists, with an interobserver consistency of 92.9%) (Table 3). The “average” is defined as the total cases independently reviewed by the two hematopathologists (254 cases) divided by two. Among the cases correctly diagnosed with CD45/SSC pattern alone, 97 cases were AML and 22 cases were ALL (Table 3). Three B-ALL and one T-ALL (total seven cases) were misinterpreted as AML, with three cases misinterpreted by both reviewers, whereas a total five AML were misinterpreted as ALL. Three B-ALL cases were considered Uncertain, while only one AML case was considered Uncertain by CD45/SSC panel analysis. With morphologic examination alone, presumptive diagnosis was correct in 95.3% of the cases, incorrect in 2.3%, and Uncertain in 2.4% (average of the two hematopathologists, with an interobserver consistency of 96.1%). When the morphologic and CD45/SSC panel data were combined, the provisional diagnosis was correct in 90.6% of the cases, incorrect in 0%, and Indeterminate in 9.4% (average of the two hematopathologists, with an interobserver consistency of 91.3%). For cases with Uncertain CD45/SSC pattern or Uncertain morphology, the provisional diagnosis was derived only from the “certain” CD45/SSC pattern or morphology, and it turned out that all the cases were correctly classified. In 109 cases (85.8%), both hematopathologists agreed on a provisional diagnosis of AML or ALL and were correct in every case (100%). In 12 cases (9.4%), one reviewer made a provisional diagnosis of Indeterminate, and the other reviewer made a provisional diagnosis of AML or ALL. The latter reviewer was correct in every case (100%), suggesting that Indeterminate cases can usually be resolved by a second opinion.

Table 3. Preliminary Classification of Acute Leukemia by CD45/SSC, Morphology, and Combined CD45/SSC and Morphology
TriagePresumptiveProvisionalFinalNa (actual number)Accuracy (%)Error (%)Uncertain/ indeterminate (%)
  • a

    Average number of cases by two reviewers.

  • b

    Total seven cases misclassified, three of which by both reviewers.

  • c

    Total three cases misclassified, two of which by both reviewers.

  • d

    Total five cases misclassified, one of which by both reviewers.

CD45/SSCAML AML97.097.0  
AML ALL3.5 [B-ALL (3), T-ALL (1)]b 13.0 
ALL ALL22.081.5  
ALL AML2.5 [AML (5)] 2.5 
Uncertain AML0.5 [AML-MC (1)]  0.5
Uncertain ALL1.5 [B-ALL (3)]  5.6
Total   94.14.31.6
MorphologyAML AML98.098.0  
AML ALL1.5 [B-ALL (2), T-ALL (1)] 5.6 
ALL ALL23.085.2  
ALL AML1.5 [AML (2)]c 1.5 
Uncertain AML0.5 [AML-M1 (1)]  0.5
Uncertain ALL2.5 [B-ALL (4)]d  9.3
Total   95.32.32.4
CD45/SSC and morphology AMLAML96.096  
 AMLALL0.0 0.0 
 ALLALL22.081.5  
 ALLAML0.0 0.0 
 IndeterminateAML4.0  4.0
 IndeterminateALL5.0  18.5
 Total  90.60.09.4

Both reviewers considered six cases (4.7%) Indeterminate. In the four cases with Uncertain morphology, CD45/SSC pattern correctly determined their lineages. Only one case was considered Uncertain by CD45/SSC pattern, but morphology alone correctly identified its lineage.

Compared to using either morphology or CD45/SSC pattern alone, combining both data to generate a provisional diagnosis resulted in a decrease in the number of provisional diagnoses and an increase in the number of Indeterminate cases (Table 3). This was not unexpected, as combining the two sources of data provided opportunity for conflict between the morphologic and flow cytometric findings (thus increasing the percentage of cases labeled Indeterminate), but also mitigated the chances of making an outright incorrect provisional diagnosis. Ultimately increasing the percentage of Indeterminate cases will result in slightly more expense in resolving those cases, but it will assure the diagnostic accuracy, which is of paramount importance.

Application of Pattern Recognition

Using both morphology and CD45/SSC pattern to establish a provisional diagnosis, we correctly predicted an average of 115 of 127 cases (90.6%). Despite this fairly high degree of accuracy, the reviewers could not make a provisional diagnosis in an average of 12 cases (9.4%). This was most often due to conflict between the morphology and CD45/SSC pattern (6.7%), but, in two cases, both the morphology and CD45/SSC pattern were considered uncertain by one reviewer. In both instances, the correct diagnosis was B-ALL, and the other reviewer made the correct provisional diagnosis.

Although we found in most of the cases that myeloblasts usually clustered in a circular or oval shape, whereas lymphoblasts tend to distribute longitudinally along the CD45 axis (Fig. 1B), the association of the blast populations with their normally maturing counterparts is more reliable for lineage determination. We encountered cases with elongated blast distribution along the CD45 axis, but with blasts showing close association with the maturing monocytes/granulocyes (Fig. 2A). They turned out to be myeloblasts. Conversely, lymphoblasts also can crowd together in an oval shape (Fig. 2B). However, the SSC of those cells never exceeded 120 U.

Figure 2.

Blasts closely associated with their mature counterparts in the CD45/SSC panel. A: Myeloblasts with elongated pattern, but intermingled with the mature monocytes; (B) lymphoblasts with oval pattern, but intermingled with the mature lymphocytes.

Six cases could not be given a provisional diagnosis and were thus labeled Indeterminate. Two were found to be de novo B-ALL; one was B-ALL transformed from chronic myelogenous leukemia; another two were T-ALL, and the last one was AML without maturation. All three B-ALLs expressed CD10, CD19, cCD79a, and two of them expressed TdT. Both of the T-ALLs expressed cCD3 and CD7, and one expressed TdT as well. The AML expressed CD13, CD33, CD117, and dim MPO. The proposed Indeterminate flow panel (Table 5) would accurately classify each of these leukemias.

Three of the cases were considered Indeterminate, because the leukemic cells were misinterpreted as myeloblasts by morphology, but lymphoblasts by the CD45/SSC pattern. One of those is a B-ALL that showed only rare circulating blasts with abundant cytoplasm and folded nuclei, giving the impression of myeloid lineage. The blasts in the CD45/SCC panel showed a typical lymphoblastic pattern. The second case is a T-ALL that displayed significant numbers of large blasts with abundant cytoplasm, irregular nuclei, and fine chromatin, although two distinct blast populations are closely associated with mature lymphocytes in the CD45/SSC panel. The last case was a B-ALL with blasts morphologically resembling myeloblasts. Interestingly, the latter two ALL cases also showed aberrant expression of myeloid lineage-associated markers CD13 and/or CD33.

In addition to acute leukemias, we in our daily practice also encounter mature B-cell and T-cell leukemia/lymphomas, such as chronic lymphocytic leukemia, leukemic variant of mantle cell lymphoma, B- and T-cell prolymphocytic leukemia, hairy cell leukemia, and adult T-cell leukemia. Because these leukemia/lymphomas express mature lymphoid markers such as surface CD3, CD4, or CD8, CD20, or surface immunoglobulins, but lack immature markers such as CD34 and TdT, it is usually not difficult to distinguish them. One pitfall is T-cell prolymphocytic leukemia, which can be morphologically lymphoblast-like and co-expresses CD4 and CD8. However, this mature T-cell leukemia is usually bright CD45-positive and intermingled with normal lymphocytes in the CD45/SSC panel.

Proposed Minimal Panels for the Diagnosis of Acute Leukemia

By examining the observed immunophenotypes of our 127 cases of acute leukemia, we identified 10 markers that were expressed in >90% of the leukemic cells (Table 4). CD7 and cCD3 were the most commonly expressed markers in T-ALL (100%). However, CD7 was aberrantly expressed in 30% of the AMLs. Cytoplasmic CD3 is considered the most lineage-specific marker for T-ALL in the WHO Classification (5) and showed no aberrant expression in AML or B-ALL. CD19, cCD79a, and HLA-DR were the most commonly expressed markers in B-ALL (100%, 95%, and 95%, respectively). CD19 and CD79a were also detected in 14% of the AMLs. CD13 and CD33 were expressed in >90% of the AMLs. In addition, all the AMLs with monocytic differentiation also expressed HLA-DR (100%). However, 60% of the T-ALLs and 14% of the B-ALLs aberrantly expressed CD13, whereas 40% of the T-ALLs and 5% of the B-ALLs aberrantly expressed CD33. CD117 was also expressed in 40% of the T-ALLs. Although several markers were expressed less frequently than the above markers, they were more specific for a particular leukemia. For example, CD10 was expressed in only 82% of the B-ALLs, but when present with CD19, cCD79a, and/or TdT it helped arrive at a final diagnosis. Although CD22 was reported to be the most specific marker for B cells (14), we usually used it only for difficult cases. TdT was expressed in 86% of the B-ALLs and 80% of the T-ALLs, and because it was very rarely (4%) positive in the AMLs, it is considered specific for ALL. In contrast, MPO was seen in only ∼66% of the AMLs. Because MPO was almost never expressed in ALL, it is considered specific for AML. On the basis of these information, we propose a minimal panel of immunological markers for the diagnosis of acute leukemias (Table 5). Stem-cell marker CD34 is commonly expressed in blasts; therefore, it is included in both AML and ALL panels. For the classification of AML with monocytic differentiations, we also use additional markers CD14 and CD64, which are more commonly expressed in acute monocytic leukemia than in the other AMLs. To diagnose acute erythroleukemia and acute megakaryoblastic leukemia, we add CD61 and glycophorin A. CD20, CD22, s-Kappa, and s-Lambda are used to distinguish B-ALL from mature B-cell leukemia/lymphomas, whereas CD3, CD4, and CD8 are added to rule out peripheral T-cell leukemia/lymphomas.

Table 5. Proposed Minimal Panels for the Diagnosis of Acute Leukemia
PanelsTubesFITCPEPerCPAPC
Screen1CD7CD33CD45CD19
Cytoplasmic1TdTnegCD45
2cTdTCD20CD45cCD3
3cMPOCD34CD45CD22
AML (optional add on)Monocytic:CD64CD14CD45HLA-DR
Other:CD61Gly-ACD45CD34
ALL (optional add on)T-cell:CD4CD8CD45CD3
B-cell:s-Kappas-LambdaCD45CD10
Indeterminate1CD64CD13CD45CD117
2CD10CD3CD45HLA-DR
Table 4. Percentage of Acute Leukemias with at Least Dim Marker Positivity
MarkerAML (n = 100)Monoa (n = 27)Other (n = 73)ALL (n = 27)B-cell (n = 22)T-cell (n = 5)
  • ∼ indicates “negative.”

  • a

    “Mono” includes M4, M4eo, AML with inv(16), M5, M5a, M5b, and AML with monocytic differentiation, NOS.

CD2108111580
CD3
CD42754181160
CD51319715560
CD730193418100
CD811740
CD10708220
CD11b35582745
CD13959296221460
CD1418468
CD19141981100
CD205973
CD33961009411540
CD34755482635980
CD5626272611540
CD64347321
CD71848584332380
CD117898192740
HLA-DR7910074899560
s-Kappa
s-Lambda45
TdT444858680
MPO666966
cCD318100
cCD79a11819520

An Algorithmic Approach

When acute leukemia is suspected after examining the morphology of peripheral blood or bone marrow aspirate, the samples will be triaged into a category of Presumptive AML, Presumptive ALL, or Uncertain and screened first with a CD45/SSC panel flow cytometry (Fig. 3). If the morphology and CD45/SSC flow cytometric pattern agree, a provisional diagnosis of AML or ALL will be rendered. Any discrepancies between the morphology and CD45/SSC pattern will be considered Indeterminate. Based on the provisional diagnosis, each specimen will be further analyzed using the corresponding AML, ALL, or Indeterminate antibody panel (Fig. 3 and Table 5).

Figure 3.

Algorithmic approach in diagnosing acute leukemia.

Although the minimal AML and ALL panels are adequate to verify the morphologic and CD45/SSC flow cytometric impressions, additional markers are often necessary to further subclassify a given neoplasm and rule out mature leukemias. To that end, our minimal panels include optional add-on tubes for AMLs with morphologic and/or CD45/SSC evidence of monocytic differentiation or erythroid/megakaryocytic differentiation. For lymphoid leukemias, a targeted add-on tube can help rule out a mature B- or T-cell neoplasm (Table 5).

In cases rendered Indeterminate after examination of morphology and CD45/SSC patterns, a 16-marker panel will be able to determine the lineage of acute leukemia in almost all the cases (Table 5). In rare cases suspected of biphenotypic acute leukemia (BAL), the diagnosis is made using a stepwise paradigm published elsewhere (15). Recognizing the importance for follow up of minimal residual diseases, we also use a small optional panel that includes the markers often aberrantly expressed by leukemia cells.

This approach can cut down the reagents from 28 antibodies to 10 antibodies in ∼90% of the new leukemia assays. Because the number of reactions is decreased and the SOP is streamlined (Fig. 4), the algorithmic approach saves ∼ $1,620 and at least 15 min of technologist time per assay. At the same time, the lineage of acute leukemia is sufficiently determined.

Figure 4.

Procedural comparison between comprehensive assay and algorithmic approach. A: Comprehensive assay, ∼2 h and 15 min to analyze one marrow specimen; (B) algorithmic approach, ∼2 h to analyze one marrow specimen. The unit of the numbers is minutes (m).

DISCUSSION

Although investigative flow cytometry will discover more interesting immunological markers for leukemia cells, clinical flow cytometry faces a dilemma to balance a limited budget and to maintain a highly accurate diagnosis of acute leukemia. With the increasing pressure on cost-effectiveness of clinical laboratories, we explored the possibility of cutting down the immunological markers that are routinely used in our flow-cytometry laboratories. Based on the CD45 antigen expression and production of granules in the leukemic cells, earlier laboratories used the CD45/SSC panel in the identification of leukemic blasts (16, 17). A retrospective analysis of the flow cytometric findings from 127 new leukemia workups in our institution demonstrated that the CD45/SSC clustering patterns of lymphoid and myeloid blasts were distinguishable from each other with high frequency. We posit an algorithm, whereby these patterns allow a specimen to be triaged into one of three antibody panels: AML, ALL, or Indeterminate. The occasional cases that are misclassified by triage should be fairly simple to detect. This algorithmic approach enabled us to diagnose acute leukemia using combined morphology and flow cytometry with minimal numbers of antibodies.

The algorithmic approach is compatible with the biology of leukemia cells in the expression of immunological markers (18). CD45 is critical for the development of hematopoietic cells (19), particularly lymphocytes, which gradually gain strong CD45 expression during maturation. In contrast, cytoplasmic granules increase with the maturation of myeloid cells. Thus, CD45/SSC panel flow cytometry could distinguish myeloblasts from lymphoblasts in the majority of cases. This approach is reliable with an interobserver agreement rate of 91.3% in our series. The stepwise approach cannot only quickly diagnose common AML and ALL, but also identify some rare biphenotypic leukemias (20). Our experience suggests that the algorithmic approach will greatly increase the efficiency and decrease the cost and turnaround time in arriving at a correct diagnosis of acute leukemia. This approach can serve as a guideline for the clinical directors of flow cytometry laboratories in managing the budget and clinical service.

The algorithmic approach is compatible with the principles of the 2006 Bethesda International Consensus Recommendations (1). Although our algorithm provides a very efficient means for the diagnosis of acute leukemia, we recognize some limitations with the current approach. An obvious problem with using clustering patterns to dictate antibody panel selection is that interpreting these patterns requires a degree of subjective analysis and experience. The patterns themselves may also be affected by instrumentation, choice of fluorochrome, the idiosyncrasies of individual antibody preparations, and by temporal adjustments made to a single analyzer. With practice and consistency in reagent selection, some of these potential difficulties can be mitigated.

Although the reduction of immunological markers undoubtedly decreases the cost of reagents, the labor and turnaround time become concerns, because our algorithm involves a triage step. To solve this issue, we designed a procedure that could streamline the analysis and compensate for the triage time. Although the specimen preparation step is the same as the comprehensive assay, the screen step is much shorter with the cytoplasmic staining set up at the beginning, it takes ∼2 h to analyze the screening panel, to make provisional diagnosis, and to analyze one additional assay. Sometimes, additional assays may not be required to determine the lineage. Approximately 15 min of the technologist time are saved per assay, which add up to 1 h per day. This saving is particularly significant for laboratories that process large number of specimens. If we consider purchasing bulk amount of fewer antibodies (discounts from suppliers), using premixed antibody cocktails (less time for setting up the assay) and employing a six-color analysis (fewer tubes), the reagents and technologist time will be saved further.

Concerns may also arise as to whether delay of additional assays by the triage will compromise the viability of leukemic cells. Our experience showed that, once processed, the cells can be analyzed within 24 h at room temperature without significant loss of viability. Because our algorithmic approach can sufficiently determine the lineage of ∼90% of acute leukemias, only rare cases such as BAL may need additional assays for diagnosis. As the markers in the Screen and Cytoplasmic panels (Table 5) are the same as in the comprehensive panel (combined Screen, Cytoplasmic, and Indeterminate panels), delay of additional assays by the triage will be compensated by effective selection of the more relevant immunological markers (Fig. 5), should a comprehensive assay is needed.

Because of the limited number of markers used in this tailored analysis, markers aberrantly expressed by leukemic cells may be missed by this approach. These aberrantly expressed markers, although not necessary for initial diagnosis, are helpful for the follow up of minimal residual diseases. Therefore, the current approach is more suitable for the initial diagnosis of acute leukemia at the smaller laboratories where the costs and turnaround time are paramount. Although the algorithmic approach is more efficient in laboratories that handle large numbers of acute leukemia daily, it does not save significant amount of time in analyzing a single specimen. It also has the disadvantage in providing less information for the new leukemia than the comprehensive assay, which tends to provide a richer academic definition of the acute leukemia.

The decision to adopt a triage algorithm would be an individual one, requiring consideration of a given flow cytometry laboratory's revenue, expenses, staffing, work flow, instrumentation, and the nature of the patient population it serves. If successfully implemented, it may serve as a creative method to reduce costs while maintaining diagnostic accuracy. Although the current algorithm is very efficient in the diagnosis of acute leukemia, this approach can be further improved by future multicenter collaborative studies to validate its consistency. The procedures should be standardized for easier use and good reproducibility. The antibodies should be standardized by the reputable suppliers, and premade useful panels may prove to increase the analytical efficiency. The number of markers devised for follow up should be balanced among the diagnostic accuracy, the costs, and the actual reimbursement of the assays.

Ancillary