Immunophenotyping is an independent factor for risk stratification in AML

Abstract

Background

Chromosomal abnormalities are one of the most important prognostic factors in acute myeloid leukemia (AML). However, only a limited number of patients have such informative chromosomal abnormalities. The prognostic value of immunophenotyping in this disease is still unclear.

Methods

Seven hundred and eighty-three newly diagnosed AML patients treated in the German SHG-AML trials in 1991 and 1996 were analyzed with a panel of 33 antibodies. Expression was correlated to overall survival, complete remission-rate, and complete remission duration, and tested in a multivariate analysis including other clinical and biological markers.

Results

With a median follow-up of 4.3 years, patients with AML blasts negative for CD9, CD11b, CD13, CD34, and CD41, or positive for CD15, CD33, CD38, CD64, and MPO had superior overall survival. This effect was associated with a significantly higher complete remission rate (CD13, CD34, CD41, and CD64) or a longer complete remission duration (CD9, CD11b, and CD64). Cox-regression analysis, including cytogenetic, morphologic, and biologic parameters showed CD9, CD13, CD34, and CD64 as independent factors for overall survival. These markers were used for a prognostic score. Patients were pooled in three groups with highly significant differences of overall survival. The prognostic relevance of this score was confirmed in patients with normal karyotype and/or in younger patients ≤ 60 years.

Conclusions

Immunophenotyping is not only helpful for diagnosis but is of independent significance for prognosis, and may be useful for risk stratification in AML patients. Cytometry Part B (Clin. Cytometry) 53B:11–19, 2003. © 2003 Wiley-Liss, Inc.

Immunophenotyping is very useful to diagnose and classify acute myeloid leukemia (AML), thereby complementing morphology, cytochemistry, and cytogenetics (1, 2). Although a large variety of clinical and biological parameters, including age over 60 years, secondary versus de novo AML, a high white-cell count, an elevated serum lactate dehydrogenase (LDH) level at presentation, and FLT3 gene mutations (3), have been claimed to correlate with response to treatment and survival in AML (2), very few of them are accepted to guide therapeutic decisions (1). Cytogenetic abnormalities represent one of the most accepted independent risk factors. However, only half of the patients have such abnormalities, which are only partially informative regarding prognostic implications. The prognostic relevance of surface antigen expression is still unclear (4). Several reports suggested a relationship between some antigens (e.g., CD7, CD9, CD11b, CD13, CD14, CD15, CD33, CD34, CD38, CD56, CD117, MPO, and TdT [terminal deoxynucleotidyl transferase]) and prognosis (4–12); however, other studies produced conflicting results.

The purpose of this study was to evaluate the prognostic significance of different immunophenotypic subgroups in a large patient cohort. In addition, the results of other prognostic features were compared within the context of such a clinical trial of adult AML patients treated with chemotherapy. Using multivariate analysis, the prognostic value of immunophenotype was determined in light of other well-known prognostic factors only in a limited number of patients (6, 7, 11–15). Here, we analyzed blast cells from 783 patients with AML with a panel of 33 antibodies.

PATIENTS, MATERIALS, AND METHODS

Patients

A total of 1426 patients with untreated AML were enrolled in the German multi-center treatment trial of the SHG-AML'91 study from August 1991 to May 1996 (488 patients) and in the SHG-AML'96 study from February 1996 to May 2000 (938 patients). The last follow up was in July 2002. Immunophenotyping was performed centrally in Erlangen, Germany in 783 patients (54.9%; 193 cases of AML'91 and 590 from AML'96 study). Sixty-seven cases were excluded from the analysis due to more than 25% CD3-positive cells, indicating a substantial proportion of non-blast cells in the analyzed sample. Chromosome analyses were performed in 660 of the remaining 716 patients on metaphases from direct preparations, as well as from 24-h and 48-h cultures of bone marrow and/or peripheral blood samples either in one of four central laboratories or at the local department of genetics. Mutation status of FLT-3 was analyzed by Dr. Thiede, Dresden, Germany, as recently published (3). White blood cells (WBC) and LDH were tested at the time of diagnosis by routine clinical chemistry.

Treatment Protocols

The treatment strategy of the SHG-AML'91 protocol consisted of a double induction therapy with two courses of DAE (daunorubicin 60 mg/m² days 4–6, cytosine arabinoside 200 mg/m² days 1–3 over 3 h, and 100 mg/m² days 4–8 as continuous infusion, VP-16 150 mg/m² days 1–3) in patients ≤ 60 years. Patients > 60 years old received two courses of DA (daunorubicin 45 mg/m² days 3–5, cytosine arabinoside 100 mg/m² days 1–7). As post-remission therapy, patients < 50 years with a matched family donor were offered allogeneic transplantation, and all other patients ≤ 60 years received MAMAC (cytosine arabinoside 2 × 1000 mg/m² days 1–5, m-amsacrine 100 mg/m² days 1–5) followed by MIHAC (cytosine arabinoside 2 × 3000 mg/m² days 1–6, mitoxantrone 10 mg/m² days 4–6) for patients ≤ 50 years or MIDAC (cytosine arabinoside 2 × 1000 mg/m² days 1–6, mitoxantrone 10 mg/m² days 4–6) for patients 51–60 years. Post-remission therapy in patients > 60 years was not specified in the protocol.

The treatment schedule of the SHG-AML'96 trial has been previously published (16). Briefly, double induction therapy was stratified according to age. Patients ≤ 60 years old received one course of MAV (mitoxantrone 10 mg/m² days 4–8, cytosine arabinoside 100 mg/m² days 1–8, VP-16 100 mg/m² days 4–8) and a second course of MAMAC. Patients over 60 years old were treated with two courses of DA. Complete remission was defined as the presence of < 5% of blast cells in a standardized bone marrow puncture after the second course of induction therapy. Only patients with a fully regenerated peripheral blood count were considered to be in complete remission.

Post-remission therapy for individuals ≤ 60 years old was priority-based and adapted according to cytogenetic risk. Patients with t(8;21) were considered “low risk,” patients with −5/del (5q), −7/del (7q), hypodiploid karyotype (except 45,X,−X and −Y), inv(3q), abnormalities of 12p or 11q; +11; +13; +21; +22; t(6;9); t(9;22); t(3;3); multiple aberrations, or treatment-related secondary AML were considered “high risk,” and all other patients as “intermediate risk.” In intermediate- and high-risk patients, an allogeneic transplantation with a family donor had the highest priority. If no HLA-identical sibling was available, a matched unrelated donor transplantation was attempted for the high-risk group. All other patients were randomized to receive a first post-remission therapy with either I-MAC (cytosine arabinoside 2 × 1000mg/m² days 1–6, mitoxantrone 10 mg/m² days 4–6) or H-MAC (cytosine arabinoside 2 × 3000mg/m² days 1–6, mitoxantrone 10 mg/m² days 4–6). Afterwards, autologous stem cell transplantation was performed in intermediate- and high-risk patients. If stem cells could not be collected or in the case of low-risk patients, a second post-remission therapy course of MAMAC was given. Patients more than 60 years usually received only one post-remission therapy course of MAMAC six weeks after complete remission.

Patients with acute promyelocytic leukemia (M3) were treated in a separate all trans retinoic acid (ATRA)-containing protocols (17). However, patients with AML M3, who were evaluated for the trial AML'96, were documented centrally for treatment, cytogenetics, and outcome. Such data are available only for a few patients from trial AML'91. In total, 41 patients with AML M3 from both trials were included in this analysis.

Immunophenotyping

Heparinized bone marrow (n = 601) or, if sufficient bone marrow was not available, peripheral blood (n = 182) samples were analyzed at the time of diagnosis before treatment was initiated. Cells were isolated by density gradient centrifugation. Cell-surface antigens were detected by standard indirect immunofluorescence as previously described (18). Antibodies used for this analysis included: CD2 (Leu-5b), CD4 (Leu-3a), CD20 (Leu-16), CD58 (AICD58), and CD64 (10.1) (BD-Biosciences, Mountain View, CA); CD3 (UCH-T1), CD9 (Alb6), CD10 (J5), CD13 (SJ1D1), CD14 (RMO52), CD15 (80H5), CD19 (J4.119), CD24 (Alb.9), CD32 (2E1), CD33 (My9), CD34 (QBEND10), CD36 (FA6-152), CD38 (T16), CD41 (P2), CD45 (J-33), CD65 (88H7), CD95 (UB2), CD117 (17F11), Glycophorin A (D2-10), and TdT (HT1,4,8,9) (Coulter Clone, Hialeah, FL); and CD56 (T199), CD61 (Y2/51), and myeloperoxidase (RMO52) (DAKO Diagnostika, Hamburg, Germany). CD11b (OKM 1), HLA-DR (L227), and isotype control antibody (10.-3.6.2.) was produced in our laboratory from the hybridoma clones CRL 8026, HB 96, and TIB 92, respectively, which were obtained from the American Type Culture Collection (ATCC-Rockville, MD). CD7 (TH-69), CD75 (EBU-141), CD96 (TH-111), and TC12 were developed in our laboratory and have been confirmed by the international Human Leukocyte Differentiation Antigens (HLDA) workshops, as published previously (18, 19). A fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse antibody detecting IgA, IgG, and IgM served as a second antibody.

During antibody incubation, polyclonal human IgG (4 mg/ml) was added to inhibit nonspecific binding to FcγRI. After washing cells four times in phosphate buffered saline (PBS) plus 1% bovine serum albumin (BSA), FITC-labeled F(ab′)₂ fragments of goat anti-mouse monoclonal antibody (MoAb) were added for 30 min at 4°C. Cells were washed again and resuspended in PBS for analysis on a flow cytometer (EPICS Elite, Profile II, or XL, Coulter Electronics, Hialeah, FL). The forward-angle light scatter was used to gate on all cells except debris. In addition, an ultraviolet (UV) microscope was used for verification when indicated. Cytoplasmic (cyCD3 and myeloperoxidase) or nuclear (terminal deoxynucleotidyl transferase) staining was performed according to standard protocols (20). The cutoff point for scoring a specimen “positive” was 25% or more antibody positive cells.

Statistical Analysis

For descriptive statistics, median and range, mean, or percentage of cases were calculated.

Overall survival and complete remission duration were analyzed for each marker separately and for other clinical and biological parameters using Kaplan-Meier curves, and differences between groups were compared using the log-rank test for univariate analysis. P < 0.05 was considered significant. Differences of marker expression (“positive” versus “negative”) or other categorical variables and complete remission-rate were evaluated with the chi-square test. The association of age, year of diagnosis, leukocyte count, and LDH at diagnosis with complete remission rate was analyzed with Mann-Whitney U test after excluding normal distribution of these factors with Lilliefors-Test. P < 0.05 was considered significant.

For multivariate analysis of prognostic factors, Cox's proportional hazard regression model was used for overall survival and remission duration. Logistic regression was used for complete remission rate. Covariates including sex, age, year of diagnosis, study protocol, French-American-British Cooperative Group (FAB) morphology, de novo versus secondary AML, leukocyte count at diagnosis, LDH, FLT-3 mutational status, risk group, and cytogenetics, were selected on the basis of significant results in the univariate analysis. A stepwise forward selection was performed and variables were added at P < 0.05 and were deleted at P > 0.10. Finally, a prognostic score was calculated including all markers, which were significant in the multivariate analysis for overall survival (CD9 negative, CD13 negative, CD34 negative, and CD64 negative). Based on the sum of favorable expression patterns, overall survival of three groups was analyzed with a score of 0, 1 + 2, and 3 + 4 using Kaplan-Meier curves and log-rank tests. All calculations were performed using the SPSS software package, version 10.0 (SPSS, Chicago, IL).

RESULTS

Patient Characteristics

We analyzed 783 samples of patients enrolled in the German multi-center treatment trial of the SHG-AML'91 and AML'96, which was 54.9% of all patients included in these protocols from August 1991 to May 2000. Because no attempts to select for cell subpopulations such as “blast cells” were allowed, 67 cases were excluded from the analysis due to more than 25% CD3-positive cells, indicating a substantial proportion of contaminating normal cells in the analyzed sample. Patient characteristics of the remaining 716 patients are listed in Table 1. Median overall survival rates were similar to 710 patients participating in the study but not submitted to our institution, with a median survival of 1.08 and 1.22 years, respectively (P = 0.763), indicating a representative subgroup in terms of survival. The median follow up of the evaluable patients was 4.28 years, ranging from 0.05 to 10.42 years.

Table 1. Patient Population

a Definition see Materials and Methods.
Patients, n	716
Sex
Male, n (% of total)	372 (52.7%)
Female, n (% of total)	339 (47.3%)
Age, median (range)	55.1 (16.4–79.3)
Disease status
De novo AMI	619 (86.5%)
Prior MDS	69 (9.7%)
Secondary AML	21 (2.9%)
WBC at diagnosis (n = 506), G/I, median (range)	24.9 (0.5–453.0)
LDH (n = 483), μmol/sl, median (range)	15.1 (0.1–120.8)
FLT-3 (n = 514)
Wild-type	339 (55.7%)
Mutated	115 (16.1%)
FAB morphology, n (% of total)
M0	32 (4.5%)
M1	126 (17.6%)
M2	228 (31.8%)
M3	41 (5.7%)
M4	148 (20.7%)
M5	111 (15.5%)
M6	16 (2.2%)
M7	4 (0.6%)
Karyotype (n = 660), n (% of total)
Abnormal karyotype	316 (47.9%)
Deletion 5	35 (5.3%)
Deletion 7	57 (8.6%)
Monosomy	33 (5%)
Trisomy 8	41 (6.2%)
Trisomy 21	14 (2.1%)
Abn. of chr. 11	27 (4.1%)
t(8; 21)	35 (5.3%)
t(15; 17)	23 (3.5%)
Inv16	37 (5.6%)
Complex aberrations	67 (10.2%)
Low risk groupa	29 (4.1%)
Intermediate risk groupa	416 (58.1%)
High risk groupa	220 (30.7%)

Relationship Between Immunophenotype and Response or Survival

A panel of 33 antibodies was used to evaluate the expression in 716 AML patients. Almost all antigens were tested in more than 80% of patients except CD38 (19.8% of patients tested), CD75 (20.0%), and CD56 (64.5%). After Ficoll separation, all cells were gated using forward-angle light scatter. A marker was considered positive when ≥ 25% of cells expressed the antigen. Survival data were available from 715 of 716 patients, and 406 reached a complete remission (56.9%) with a median continuous complete remission (CCR) of 2.32 years. An estimated 44.4% of all patients achieving a complete remission remained in CCR after five years. Median overall survival was 1.08 years with an estimated two- and five-year survival of 36.8% and 26.4%, respectively.

The overall survival was significantly worse in patients found positive for CD9, CD11b, CD13, CD34, and CD41, but superior in cases expressing CD15, CD33, CD38, CD64, and MPO (Table 2 and Fig. 1). Additional antigens with a trend toward better or worse survival (P between 0.05 and 0.1) are listed in Table 2. A lower complete remission rate was found for patients expressing CD10, CD13, CD34, CD41, CD61, CD117, and HLA-DR. A higher complete remission rate correlated significantly with the expression of CD32, CD56, CD64, and CD65. CCR was shorter in cases positive for CD4, CD9, and CD11b, but significantly longer with CD64 expression (Table 2).

Table 2. Summary of Univariate Analysis

Antigen	OS (P) log rank	CR rate χ^2 (P)	CCR (P) log rank	Favorable when
* P < 0.05; ** P < 0.01; OS, overall survival; CR, complete remission.
CD4	0.0584		0.0489*	Negative
CD9	0.0004**	0.083	0.0401*	Negative
CD10	0.0633	0.034*		Negative
CD11b	0.0042**	0.054	0.0240*	Negative
CD13	<0.0001**	<0.001**	0.0605	Negative
CD15	0.0073**	0.052		Positive
CD32		0.032*		Positive
CD33	0.0487*			Positive
CD34	0.0009**	<0.001**		Negative
CD38	0.0385*			Positive
CD41	0.0305*	0.003**		Negative
CD56	0.0786	0.006**		Positive
CD58	0.0542			Positive
CD61	0.0784	0.030*		Negative
CD64	0.0003**	0.001**	0.0033**	Positive
CD65		0.048*	0.0856	Positive
CD75		0.059		Positive
CD117		0.037*		Negative
HLA DR	0.0917	0.019*		Negative
TC12			0.0556	Positive
MPO-7	0.0094**	0.076		Positive

Figure 1. Influence of antigen expression on overall survival. Correlation of the expression of CD9 (A), CD13 (B), CD34 (C), and CD64 (D) with overall survival of AML patients (Kaplan-Meier curves). Statistical comparisons between groups based on log-rank tests (P-values shown in the figures). Cases with ≥ 25% positive cells (lower curves) were compared to cases with < 25% (upper curves); numbers of patients in each group are indicated.

Influence of Other Parameters on Response or Survival

Other clinical and biological parameters, including sex, age, year of diagnosis, study protocol, FAB morphology, de novo verses secondary AML, leukocyte count at diagnosis, LDH, FLT-3 mutational status, risk group, and cytogenetics, were evaluated in a univariate analysis for the influence on overall survival, complete remission-rate, or CCR (Table 3). The definition of high risk, intermediate, and low risk groups, based mainly on cytogenetic aberrations, is described in the Material and Methods.

Table 3. Significance of Covariates

Clinic and cytogenetics	OS (P) log rank	CR rate χ^2 (P)	CCR (P) log rank	Favorable when
* P < 0.05; ** P < 0.01; n.s., not significant; OS, overall survival; CR, complete remission. a Tested as single covariate in cox regression for overall survival, CCR, and with Mann-Whitney U test for complete remission rate.
Sex				n.s.
Agea	<0.001**	<0.001**	<0.001**	Younger
Year of diagnosisa		0.003**		Earlier
AML '91 versus '96		0.003**		AML '91
FAB	0.0001**	0.001**	0.0002**	M3, no M0 or M1
De novo versus s	<0.0001**	<0.001**		De novo
WBCa				n.s.
LDHa				n.s.
FLT-3			0.0702	Trend for wild-type
Risk group	<0.001**	<0.001**		Low > stand. > high
Deletion 5	<0.001**	<0.001**	<0.001**	No deletion 5
Deletion 7	0.0008**	0.003**	0.0187*	No deletion 7
Monosomy	0.0025**	0.008**	0.0432*	No monosomy
Trisomy 8			0.0293*	No trisomy 8
Trisomy 21				n.s.
t(8; 21)	0.0005**	<0.001**		t(8; 21)
t(15; 17)	0.0026**	0.012**	0.0083**	t(15; 17)
Inversion 16	0.0271*	0.007**		Inversion 16
Abn. of chr. 11				n.s.
Complex aberr.	0.0145*	0.030*		No compl. Ab.

A significantly shorter overall survival was found for older age, FAB M0 or M1, secondary AML, deletion 5, deletion 7, monosomy, and complex aberrations. A better overall survival was found for FAB M3, de novo AML, t(8;21), t(15;17), and inv(16) (Table 3). Complete remission rate was significantly worse with older age, year of diagnosis, study AML'96 versus AML'91, FAB M1, secondary AML, deletion 5, deletion 7, monosomy, and complex aberrations, and better for FAB M3, de novo AML, t(8;21), t(15;17), and inv(16) (Table 3). The somewhat inferior outcome of patients treated in study protocol AML'96 may be explained by the higher median age of 57.9 years compared to 50.3 years in AML'91 (P < 0.001). Complete remission duration decreased significantly with older age, FAB M0, deletion 5, deletion 7, monosomy, and trisomy 8, and increased with FAB M3 and t(15;17) (Table 3). Patients without mutations of FLT-3 showed a strong trend towards a longer CCR (P = 0.0702). Prognostic significance of risk groups was confirmed for overall survival, complete remission-rate, and CCR.

Multivariate Analysis

Parameters influencing overall survival and CCR were analyzed using Cox's proportional hazard regression model. Logistic regression was used for complete remission rate. Each marker was tested together with other clinical and biological parameters that were significant in the univariate analysis (Table 3). Covariates used for overall survival included age, FAB morphology, de novo versus secondary AML, risk group, and cytogenetics. For complete remission rate, the covariates age, year of diagnosis, study protocol, FAB morphology, de novo versus secondary AML, risk group, and cytogenetics were used, while for CCR, the parameters age, FAB morphology, and cytogenetics were selected.

Five out of nine markers that had a significant influence on overall survival in the univariate analysis remained significant in the multivariate analysis. In detail, overall survival was influenced by expression of CD9 (P = 0.007), CD13 (P < 0.001), CD33 (P = 0.019), CD34 (P = 0.001), and CD64 (P = 0.004) (Table 4). Multivariate analysis revealed the influence of CD13 (P < 0.001), CD34 (P < 0.001), CD41 (P = 0.012), CD56 (P = 0.019), CD64 (P = 0.007), and CD117 (P = 0.005) for complete remission rate. Only CD64 remained significant in multivariate analysis for CCR (P = 0.012) (Table 4).

Table 4. Summary of Multivariate Analysis

Antigen	OS (P) cox	CR rate (P) log. reg.	CCR (P) cox	Favorable when
* P < 0.05; ** P < 0.01; n.s., not significant; OS, overall survival; CR, complete remission; log. reg., logistic regression; cox, cox regression.
CD4			n.s.
CD9	0.007**		n.s.	Negative
CD10		n.s.
CD11b	n.s.		n.s.
CD13	<0.001**	<0.001**		Negative
CD15	n.s.
CD32		n.s.
CD33	0.019*			Positive
CD34	0.001**	<0.001**		Negative
CD41	n.s.	0.012*		Negative
CD56		0.019*		Positive
CD61		n.s.
CD64	0.004**	0.007**	0.012*	Positive
CD65		n.s.
CD117		0.005**		Negative
HLA DR		n.s.
MPO-7	n.s.

Prognostic Score

CD9, CD13, CD33, CD34, and CD64 remained significant in the multivariate analysis for overall survival. Using these markers together as covariates in a Cox multivariate analysis, CD9 (P < 0.001), CD13 (P = 0.003), CD34 (P = 0.037), and CD64 (P = 0.001) remained significant, indicating an independent influence on overall survival. Therefore, we evaluated a prognostic score with one point each for CD9 negative, CD13 negative, CD34 negative, and CD64 negative. Patients were pooled in three groups with a score of 0, 1 + 2, and 3+4, and highly significant differences of overall survival were found using log-rank tests (Fig. 2A). The estimated five-year survival rates of these groups were 47.8%, 27.3%, and 7.97%, respectively. Because patients with FAB-M3 and older patients were treated separately in both trial AML'91 and AML'96, the score was tested for its usefulness in the group of patients ≤ 60 years old, excluding M3 AML. As shown in Figure 2B, significant differences between all three groups were found. Risk stratification in AML'96 was based on cytogenetic abnormalities, with the largest group of 52.1% having a normal karyotype. Also for these patients, the prognostic score remained highly significant for overall survival (Fig. 2C) and even for patients ≤ 60 years with a normal karyotype (Fig. 2D).

Figure 2. Score system stratifying AML patients. Overall survival of AML patients according to a score based on CD9, CD13, CD34, and CD64 expression (Kaplan-Meier curves). Favorable patients with negativity (< 25% positive cells) for three or four markers (upper curves) are compared to intermediate patients with one or two negative markers (intermediate curves) and unfavorable patients (lower curves) with one or no negative markers. P-values of log-rank tests comparing group 0 versus 1 + 2 and 1 + 2 versus 3 + 4 are indicated. P-values of group 0 versus 3 + 4 were ≤ 0.0001 for all four plots. Significant differences were found for all patients (A) and also in younger patients excluding M3 (B), as well as in patients without cytogenetic abnormalities (C). Significant differences could even be found in the subgroup of younger patients with a normal karyotype (D).

A bone marrow specimen was available in 76.8% of patients. In all other patients, peripheral blood was analyzed. In order to exclude that the kind of material affects the risk stratification, the prognostic score was evaluated for patients with bone marrow samples only. Again, significant differences of the overall survival were found between all three groups, with P-values of ≤ 0.0003. In patients with peripheral blood samples only, due to the low number of patients, a significant survival difference was found only between the group with score of 0 versus a score of 1 + 2 (P = 0.0042).

To exclude that the interpretation of the data might be influenced by the arbitrary chosen cut-off value of 25% for “positivity” of antigen expression, data were reanalyzed using a cut-off point of 15%. In the univariate analysis of the influence on overall survival, only CD38 and CD41 did not reach significance. Cox regression analysis confirmed CD9 (P < 0.001), CD13 (P < 0.001), CD33 (P = 0.034), CD34 (P = 0.004), and CD64 (P = 0.000) as independent prognostic factors. Evaluation of the prognostic score showed highly significant differences between the three groups. Including all patients, differences of overall survival of the group with score of 0 versus 1 + 2, and score 1 + 2 versus 3 + 4 were significant, with P < 0.0001 and P = 0.0013, respectively. In patients with a normal karyotype, all three groups remained significantly different, with P < 0.01.

DISCUSSION

Approaches to improve AML therapy have often focused on stratifying patients for intensity of post-induction treatment according to defined risk categories based mainly on genetic criteria (2, 21). However, about half of the patients have no specific chromosomal abnormalities detectable, and additional prognostic factors are required to optimize the use of presently available therapies (22). Our retrospective flow-cytometric analysis of more than 700 patients with AML with a median follow up exceeding four years suggests several prognostically relevant markers for survival, response, and response duration. In the past, the prognostic implications of surface antigen expression in AML have been controversial. The comparability of the results may be hampered by methodological differences in the detection of antigen expression, as well as by differences in the patient populations studied and treatment regimens administered (4, 23). For example, using antibodies recognizing varying epitopes may lead to discordant results, as shown for CD34 (24). In addition, most studies have evaluated only a small number of antigens or often included less than 100 patients. Notably, in childhood, AML immunophenotyping was reported to generally have only little prognostic impact, but the relevance of marker analysis may be different in the adult population (25). Currently, only three other studies with more than 400 adult patients have been reported (26–28).

In our study, expression of CD9, CD13, CD34, and CD64 were the most important factors for survival, even after adjusting for other clinical and biological parameters such as cytogenetics, morphology, or age. Our proposed prognostic score based on the above cited markers allows us to differentiate three groups with highly significant survival differences, even for the subgroup of patients with a normal karyotype. In a univariate analysis, we found a correlation of marker expression and survival for 10 out of 33 markers. This is consistent with reports for individual markers such as CD9 (6, 29), CD11b (6, 30, 31), CD13 (10, 31), CD15 (10, 13), CD33 (7, 9), CD34 (8, 9, 13, 14, 32–34), CD38 (11), CD41 (29), and myeloperoxidase (12). However, some studies could not find an influence on outcome (28, 35–37), or demonstrated a prognostic relevance only for a score based on several myeloid markers (15). Results opposite to our findings were rarely reported, as for CD11b (38). A meta-analysis might help to further clarify the prognostic relevance of these markers, as recently provided for CD34 (39). To our knowledge, the role of CD64, which is strongly correlated to response to treatment and outcome in our study, has not been evaluated in AML before. This might be due to the fact that CD64 antibodies are only recently being used more frequently for AML immunophenotyping. Concerning lymphoid markers, such as CD2, CD7, CD19, and TdT, some reports including ours have not found statistically significant differences (8, 32, 40, 41), in contrast to other reports (5, 9, 10, 27, 42, 43). However, this may be due to the significance level and that subgroups of AML might not be well characterized by a single marker but rather by a combination of different markers. In fact, we could find a significantly better prognosis for a subgroup of patients with coexpression of CD7 and TC12 (data not shown). On the other hand, antigen expression may have a varying prognostic impact in distinct AML-subgroups, which might explain controversial results concerning CD56 expression. In the subgroup of patients with APL (44) or t(8;21) translocation (45), CD56 positivity was a negative prognostic factor, whereas in other subgroups, the data are conflicting (8, 43, 46–48), including our finding of a significant increased complete remission rate in CD56-positive patients.

In our study, we preferred bone marrow samples for flow cytometry. However, because patients with failed bone marrow aspiration (“dry tap”) may represent a relevant subgroup in AML, we accepted peripheral blood samples in these cases because earlier reports found no differences of surface marker expression from both sources (49). Our proposed prognostic score remained valid for patients with bone marrow samples only. Due to the low number of patients with peripheral blood samples, only two groups remained significantly different. Because we analyzed blasts only from one material per patient, we cannot exclude differences in the prognostic implications using peripheral blood exclusively. This is probably of minor relevance, because a bone marrow biopsy is performed anyway for other reasons in the majority of AML patients (49).

We used a threshold of 25% for considering a marker positive; in most studies this threshold is set to 15%–25%. However, analyzing our data with a cut-off of 15% did not show relevant changes. Eight instead of 10 positive markers were significant in the univariate analysis, but results for the multivariate analysis and for evaluation of the prognostic score did not show any difference. A way to circumvent the arbitrary choice of a distinct positivity threshold may be a retrospective cut-off point for each individual marker (30).

Although immunophenotyping is widely regarded as contributing little to prognostic evaluation in AML, we could identify several independent prognostic factors regarding the outcome in patients with adult AML. In fact, it was possible to even develop a prognostic score allowing stratification of patients according to their risk. Thus, it may be premature to rely solely on genetic findings and disregard the generally very stable protein expression on the cell surface that defines a certain subtype. Given the heterogeneity in surface marker expression in adult AML, current classification of AML subgroups, based on morphology and cytogenetics, may not yet be considered to mirror the underlying biology of the disease as achieved in lymphoma classifications.

In conclusion, using all available information may help to design innovative therapies to improve outcome in the poorer risk groups and to diminish toxicity for patients with a better prognosis.

Acknowledgements

We thank the following other members of the German SHG-AML study group who provided us with patient samples for immunophenotyping: F. Hirsch (Klinikum Offenburg), M. Wilhelm (Klinikum d. Universität Würzburg), M. Sandmann (Klinikum St. Antonius Wuppertal), D. Huhn (Virchow-Klinikum Berlin), T. Geer (Diakonie-KH Schwäbisch-Hall), J. Kaesberger (Diakonissen-KH Stuttgart), T. Wagner (Universitätsklinikum Lübeck), M.R. Clemens (Mutterhaus d. Borromaerinnen Trier), K.-P. Schalk (St. Vincent-KH Limburg/Lahn), A. Neubauer (Klinikum d. Philipps-Universität Marburg), H. Pohlmann (Städtisches KH München-Harlaching), K.-H. Pflüger (Evang. Diakonissenanstalt Bremen), H. Dürk (St. Marien-Hospital Hamm), J. Labenz (Ev. Jung-Stilling-KH Siegen), R. Engberding (Stadtkrankenhaus Wolfsburg), Dr. M. Sandmann (Kliniken St. Antonius, Wuppertal), A.A. Fauser (Klinik f. Hämatologie/Onkologie und KMT Idar-Oberstein), F. Fiedler (Klinikum Chemnitz gGmbH), J.G. Saal (Malteser Krankenhaus Flensburg), F. Marquard (Allgemeines KH Celle), H. Schmidt (Kreiskrankenhaus Hameln), R. Schwertfeger (Deutsche Klinik f. Diagnostik GmbH Wiesbaden), and H.-H. Heidtmann (St.Joseph-Hospital Bremerhaven). We also acknowledge the excellent technical assistance of Heike Kaul, Tanja Schönmetzler, Susanne Jaeckel, Irene Welsch, and Bliss Cronquist.