Professor Dr Ursula Creutzig Klinik und Poliklinik für Kinderheilkunde (Pädiatrische Hämatologie/Onkologie), Albert-Schweitzer-Str. 33, D-48129 Münster, Germany. e-mail: firstname.lastname@example.org
To define paediatric AML patients with a favourable outcome in order to design a risk-adapted therapy, we analysed 489 children under 17 years of age treated similarly in studies AML-BFM 83 and 87. 369 patients (75.4%) achieved remission. Estimated probabilities of survival, event-free survival (EFS) and disease-free survival (DFS) at 5 years were 50% (SE 2%), 43% (SE 2%) and 58% (SE 3%), respectively. Multivariate analysis revealed bone marrow blasts on day 15, morphologically defined risk groups and hyperleucocytosis to be of prognostic value. EFS at 5 years estimated for patients with 5% and >5% blasts on day 15 were 56% (SE 3%) v 27% (SE 4%); for the favourable morphological subgroups (M1/M2 with Auer rods, M3 and M4eo) it was 60% (SE 4%) compared with other patients (33%, SE 3%), P (Kaplan-Meier) = 0.0001 each. Hyperleucocytosis proved to be an independent prognostic factor, indicating a high risk, especially for early failure. The specific karyotypes t(8;21), t(15;17) and inv16 were closely related to the favourable morphology and outcome was in the same range. We conclude that for the definition of a standard-risk group a combination of morphological and response criteria may be sufficient. The standard-risk group defined by favourable morphology and a blast cell reduction on day 15 (not required for M3) comprises 31% of all patients, P survival, pEFS and pDFS at 5 years were 73% (SE 4%), 68% (SE 5%) and 76% (SE 4%), respectively.
The prognosis for children with AML has improved considerably during the last 15 years (Hann et al, 1997; Lie, 1995; Woods et al, 1996) with nearly 50% long-term survivors. This has been achieved mainly by a more intensified chemotherapy and better supportive care. Nevertheless, with the current therapy protocols only half of the patients can be cured. Thus, more intensive treatment (including bone marrow transplantation [BMT]) for children with a high risk for non-response and relapse is warranted. However, this should be achieved without increasing the toxicity for children who can be cured with standard AML treatment. Currently, the BMT-related mortality rate after allogeneic HLA-identical BMT is approx. 10–20% (Gorin et al, 1996; Zittoun et al, 1995) and is about 30% after unrelated BMT in children and adults (Appelbaum, 1997).
The aim of our analysis was to identify patients with a favourable prognosis (standard-risk [SR] group) in order to apply a risk-adapted therapy, thus reducing the treatment-related mortality and morbidity.
From study AML-BFM 83 (Creutzig et al, 1990) we defined our low-risk patients by morphological criteria including FAB types with granulocytic differentiation and specific additional features such as Auer rods and eosinophils. In study AML-BFM 87 the prognostic significance of blast cell reduction on day 15 was established. Therefore we re-evaluated our data to define the significance of ‘blast cell count on day 15’ for the patient group with favourable morphological criteria.
Currently, cytogenetic and molecular findings are considered important prognostic factors in AML which may influence therapeutic decisions (Martinez Climent, 1997). Based on our experiences and those of others, precise correlations between well-defined morphological and cytogenetic subtypes were identified (Creutzig et al, 1995; Nakamura et al, 1997). Centrally reviewed FAB typing was done for all patients in our studies, whereas karyotyping was not performed for the entire patient group. Our risk stratification, based on the retrospective analysis of studies AML-BFM 83 and 87, can be used for all patients shortly after induction therapy.
PATIENTS AND METHODS
Between December 1982 and October 1992, 489 previously untreated children with AML under 17 years of age were enrolled in the multicentre studies AML-BFM 83 and 87. Children with prior malignancies, with Down's syndrome, myelosarcoma and myelodysplastic syndromes were excluded as were children with more than 14 d of pretreatment (e.g. non-AML-specific therapy). All parents/patients gave their informed written consent.
The initial diagnosis of AML and its subtypes was determined according to the FAB classification (Bennett et al, 1985a, b, 1991). In addition to the Pappenheim-stained bone marrow (BM) and blood smears the following cytochemical stainings were performed: periodic acid-Schiff (PAS), myeloperoxidase (POX), alpha-naphtyl-acetate-esterase (ANAE) and acid phosphatase. All smears were routinely investigated at the University Children's Hospital in Münster and were reviewed by a panel of haematologists including one or two external investigators (T. Büchner, H. Löffler). The diagnoses of M0 and M7 subtypes always required confirmation by immunological methods (Bennett et al, 1985a, 1991). Day 15 bone marrow aspirates from 392/489 patients (80%) underwent central review. Missing values occurred mainly in early death patients. Leukaemic and non-leukaemic precursor cells were differentiated from the residual cells of the generally hypocellular bone marrow.
The treatment strategies of the AML-BFM 83 and 87 protocols have been described (Creutzig et al, 1990, 1993). In both studies therapy started with an 8 d induction course comprising cytosine arabinoside (Ara-C), daunorubicin and etoposide, followed by an 8-week consolidation (study 83) or a 6-week consolidation (study 87) with seven different drugs. Two blocks of late intensification with HD-Ara-C (6 × 3 g/m2) and etoposide were added in study 87 (Fig 1). Maintenance therapy consisted of thioguanine daily and Ara-C for 4 d every 4 weeks up to a total of 24 months (study 83) or 18 months (study 87). In the latter study, 94/230 responding patients did not receive cranial irradiation. Allogeneic BMT in first remission was recommended for children in the high-risk group in study 87 only (Creutzig et al, 1990).
Immunological classification of cell surface markers was performed in 123/182 children (68%) in study 83 and in 267/307 children (87%) in study 87 at the central reference laboratory of the AML-BFM studies (W. D. Ludwig, Universitätsklinik Charité, Berlin-Buch). Results and the clinical significance of immunophenotyping in study 87 have been published (Creutzig et al, 1995).
Initial bone marrow samples were karyotyped in Giessen by J. Harbott and F. Lampert. For methods see Creutzig et al (1995). Results of karyotyping were available in 59 patients in study AML-BFM 83 and in 146 patients in study 87.
Parameters for assessing outcome were the response rate (rate of patients achieving CR), the event-free survival (EFS) and the disease-free survival (DFS). EFS was calculated from the date of diagnosis to the first event (relapse or death of any cause) or date of last follow-up, and DFS was defined as the time from CR until the first event or date of last follow-up. Survival curves, standard errors (SE) and tests for differences in EFS between subgroups were calculated by standard methods (Kaplan & Meier, 1958; Kalbfleisch & Prentice, 1982). The Kaplan-Meier test was used to compare point estimates of survival rates (Goldman, 1991). Cox regression (Cox, 1972) was used for multivariate analysis. Cranial irradiation and BMT were included as time-dependent co-variables. All tests were descriptive and explorative. Calculations were performed with the SAS 6.12 program (SAS Institute, Cary, N.C.).
Follow-up was in median 7.2 years (9.8 years in study 83 and 5.7 years in study 87). Data were updated on 1 April 1997.
Early death: death before or during the first 6 weeks after starting treatment. Complete remission (CR): <5% blasts in the bone marrow, no evidence of disease at any other site and haematological recovery according to the Cancer and Leukaemia Group criteria (Cheson et al, 1990). Non-response (NR): no CR after consolidation treatment. CNS leukaemia: presence of >10 cells/μl (blasts confirmed) in the cerebrospinal fluid or non-haemorrhagic intracerebral infiltrates.
1Table I shows the initial clinical data including the FAB subtypes of the 489 patients of studies AML-BFM 83 and 87 (combined and separately). The higher incidence of the subtypes M0 and M7 in study 87 compared with study 83 was due to the better definition of these subtypes since 1985 (Bennett et al, 1985a, 1991; Lee et al, 1987) and to better diagnostic methods (especially immunophenotyping) in study 87. The shift in the percentage of FAB type M1 to M2 in study 87 was caused by a more precise definition of these subtypes since 1985 (Bennett et al, 1985b). Cytogenetic results were based mainly on study 87 (for more details see Creutzig et al, 1995). Initial patient data were similar in patients with or without karyotyping (data not shown).
Table 1. Table I. Initial patient data of studies AML-BFM 83 and 87. * Percentage of patients with data, no information about Auer rods in eight patients of study AML-BFM 83 and four patients of study AML-BFM 87.† P (χ2) < 0.05.
The overall results for each study and both studies combined are presented in 2Table II. 77% of the children (early deaths before starting AML chemotherapy excluded) achieved CR, 50% were still alive after 5 years. The estimated probabilities for EFS (pEFS) and DFS (pDFS) at 5 years were 43% (SE 2%) and 58% (SE 3%). The differences between study 83 and 87 were not significant (P = 0.18 each, Kaplan-Meier test).
Table 2. Table II. Results in studies AML-BFM 83 and 87. Abbreviations: CR, complete remission; CCR, continuous complete remission; Ifu, lost to follow-up.* Including nine and four early deaths before starting therapy of studies AML-BFM 83 and AML-BFM 87, respectively.† Lfu after 0.7–11.5 (median 8.7) years.‡ At 5 years.
Regarding prognostic factors for CR rate, pEFS and pDFS, the best results have been achieved in the subgroups M1/M2 with Auer rods and M4eo (morphological SR Table III). Although patients with FAB subtype M3 showed a high risk of early death (5/20 patients = 25%), only few patients showed non-response (one patient) or relapsed (three patients) (Table IV).
Table 3. Table III. Results of different risk groups in studies AML-BFM 83 and 87 (M3 excluded). * Early deaths after day 15.
Table 4. Table IV. Results for cytogenetic groups and FAB M3 in studies AML-BFM 83 and 87 (M3 included). * Definition of favourable cytogenetics: t(8;21), t(15;17), inv16.
BM blast cell count on day 15
CR rate and pDFS were significantly better for children with a substantial blast cell reduction on day 15 (5% blasts) compared with those with >5% blasts. In contrast, M3 patients often presented with residual blasts on day 15 (13/23 = 57%); however, this did not translate into a higher NR or relapse rate (Table IV). Therefore in FAB M3 patients ‘BM blast cell count on day 15’ was not considered a risk factor.
A white blood cell (WBC) count 100 × 109/l (hyperleucocytosis) indicated a high risk for initial failure with an early death and non-response rate of 50% (both studies, Table III). The relapse rate was also high (23/48 CR patients, pDFS 46%, SE 7%); however, it was not significantly different from patients with WBC count < 100 × 109/l (pDFS 59%, SE 3%, P log rank 0.095; Table III).
Infants <1 year of age and children aged 1–2 years had a similar prognosis (pEFS at 5 years, age < 1 year v age 1–2 years: 34%, SE 7% v 23%, SE 6%, n.s.). However, children under 2 years of age had an inferior outcome compared with older children.
Patients with the favourable cytogenetic subtypes, t(8;21), t(15;17) and inv16, had a significantly better outcome compared with patients with other normal or abnormal karyotypes (Table III). Normal and abnormal karyotypes were combined, because there was no significant difference in pEFS for patients with normal karyotypes compared with unfavourable aberrations (pEFS at 5 years, normal v other karyotypes 44%, SE 7% v 31%, SE 5%, P = 0.11). Prognosis for patients without cytogenetic data was similar compared with patients with karyotyping (data not shown).
Expression of lymphoid-, progenitor- and most myeloid-associated antigens had no influence on prognosis (data not shown; Creutzig et al, 1995).
Table 5. Table V. Stepwise Cox regression analyses for EFS (M3 excluded).
Table 6. Table VI. Cox regression analyses for EFS (risk groups defined by morphological critera*, FAB M3 excluded). * Standard-risk group defined by morphological criteria: FAB M1/M2 with Auer rods, FAB M3, FAB M4eo, high-risk group: all others.† For DFS, P value for HR patients 0.008, for SR patients 0.0001.
The aim of our analysis was to select a favourable subgroup of patients before or during initial treatment who may be a target group for less intensive therapy. The patients with FAB M3 were excluded from the first calculations, because these patients were unique regarding their particular risk for early death due to bleeding complications.
The initial features suspected to have an impact on prognosis (FAB type, WBC count, liver or spleen enlargement, involvement of CNS or lymph nodes, bleeding, and age) were included in a stepwise Cox regression analysis (Table V). FAB subtype M7 and hyperleucocytosis were predictive of an unfavourable outcome. M1/M2 with Auer rods and M4eo were favourable prognostic features. The treatment factor cranial irradiation, which had shown a favourable effect on survival, had an additional influence on outcome (P = 0.013), whereas allogeneic BMT in first CR did not (P = 0.38).
In a second step all prognostically favourable morphological factors were combined in order to define a standard-risk group on the basis of morphology (SRmorph). A Cox analysis of the SRmorph. was performed on all previously mentioned factors and additionally the variable for initial response ‘BM blasts on day 15’ was included (Table VI). The latter had not been included in the first step because: (1) it was not available at diagnosis, (2) some data were missing, especially in patients with early treatment failure, and calculations excluding patients with incomplete data sets would have introduced a bias. The Cox analysis of the SRmorph. group was aimed at identifying those factors, including response to therapy, that adversely affected the prognosis within this particular group. Patients with such risk factors should be excluded from the SR group or should be shifted to the high risk (HR) group later on. It could be shown that a BM blast cell count >5% on day 15 was of additional impact, with a risk ratio of 3.9 (Table VI). Hyperleucocytosis was also found to be an adverse factor for EFS in the Cox regression.
Cox analysis for the total group of patients revealed BM blasts on day 15 > 5%, and hyperleucocytosis to predict an unfavourable outcome (risk ratio: BM blasts day 15 = 2.3, hyperleucocytosis = 1.5). Hyperleucocytosis was more frequent in patients with unfavourable morphological subtypes compared with favourable subtypes (70/300 patients = 23% v 26/189 patients = 14%, Pχ2 = 0.04).
Definition of risk groups
Based on these results, a SR and HR risk group was defined including morphological features and BM blasts ±5% on day 15. The SR group comprised 31% of all patients with the following features: M1/M2 with Auer rods, M3 and M4eo and blast cell reduction on day 15 to 5% (not required for M3). FAB M3 was included in the morphologically defined SR group because outcome of patients was good once the initial risk period for early death mainly due to bleeding complications had passed. All other FAB types belonged to the HR group.
For the SR group the estimated probabilities for survival, EFS and DFS at 5 years were 73% (SE 4%), 68% (SE 5%) and 76% (SE 4%), respectively, and for HR patients: 39% (SE 3%), 33% (SE 3%) and 47% (SE 3%). Fig 2 shows outcome of SRmorph. patients with > 5% BM blasts on day 15 to be similar to HR patients defined by morphological features only.
Hyperleucocytosis was not included in our risk definition. This parameter indicates a high risk of early death (mainly by leucostasis and haemorrhage) and non-response. For SR patients the additional risk for initial failure by hyperleucocytosis was based on small patient numbers (5/22 patients). However, after achieving remission the prognosis was comparable to others in the SR group (WBC count <100 × 109/l: pDFS = 77% (SE 4%) compared with WBC count 100 × 109/l: pDFS = 71% (SE 11%), P = 0.39).
Seventy-four of 339 patients (22%) in the HR group initially presented with hyperleucocytosis and have had an extremely poor prognosis, only 10/74 patients (13.5%) were in CCR (pEFS at 5 years: WBC <100 × 109/l v WBC 100 × 109/l = 48% (SE 3%) v 23% (SE 4%), pDFS = 49% (SE 4%) v 32% (SE 8%), P = 0.02). And children of the HR group with hyperleucocytosis and blast cell persistence on day 15 fared even worse (2/21 patients alive).
Cytogenetics were not included in the first analysis, due to the low number of cytogenetic data available. With multivariate analysis, including the defined risk groups and cytogenetics, karyotypes did not reveal an additional prognostic significance (P (χ2) = 0.2). However, a high correlation between cytogenetics and the risk groups was found (Table VII). 43/54 (80%) patients with favourable cytogenetics were assigned to the SR group and 125/151 (83%) patients with unfavourable karyotype showed HR criteria.
Table 7. Table VII. Correlation between risk groups as defined* and karyotypes. * Standard-risk group: FAB M1/M2 with Auer rods, FAB M3, FAB M4eo, in addition blasts in the BM on day 15 5% (except for M3); high-risk group: all others.
On the other hand, 26/69 (38%) SR patients presented with unfavourable karyotypes and 11/136 (8%) of HR patients with favourable karyotypes. Interestingly, both groups have had a good prognosis (Table VII).
Biologically, AML is a very heterogenous disease, not only morphologically but also by immunophenotypic and cytogenetic features. At least some entities of AML are characterized by a high correlation between morphology, immunophenotypic features and karyotypes (Kuss et al, 1994). The three subgroups of AML presenting with characteristic cellular features, M1, M2 with Auer rods, M3 and M4eo are associated with a good prognosis. Children with these subtypes have a low early death rate (except M3), a high response rate and few relapses. Recently, outcome in patients with M3 could be improved by treatment with all-trans retinoic acid (ATRA), reducing the early death and relapse rate (Degos et al, 1995; Fenaux, 1994). In study 83 and in adults the prognostic significance of Auer rods and eosinophils was found (Haferlach et al, 1993; Mertelsmann et al, 1980; Ritter et al, 1989; Creutzig et al, 1990). Other investigators reported an association of FAB type M0 and M7 subtypes with a worse prognosis in children and adults (Lee et al, 1987; Ribeiro et al, 1993; Cadwell et al, 1993).
Analyses of the MRC AML10 trial identified a good-risk group comprising 28% of children with a 7-year survival of 78%, for whom BMT in first CR was not mandatory (Stevens et al, 1998). Favourable karyotypes and M3 were predictive for this good-risk group, which might not be very different from our SR group, because of the high correlation between favourable morphology and favourable cytogenetics. Outcome in both studies was similar.
According to the results of studies AML-BFM 83 and 87 we could define a standard-risk group with >70% of patients being alive after 5 years. This group comprised one third of all patients. The definition of standard-risk patients was based on the criteria of morphology and blast cell reduction on day 15. Sackman Muriel et al (1996) could confirm this definition by using nearly the same therapy schedule.
As risk factors are related to the intensity of treatment, patients with pre-existent disease, such as Down's syndrome or secondary AML, and who were treated differently, were not included in our analysis. Our study population was relatively homogenous and treated similarly with only a few modifications in study AML-BFM 87 compared with study 83.
Including the data of study 87, our analysis revealed the significance of blast cell reduction evaluated by BM aspirate on day 15 for the response and relapse rate, indicating an increased relative risk of failure (3.9 for SR patients). This finding was supported by the Children's Cancer Study Group (CCG) who reported a significantly shorter survival in children with more than 5% blasts on day 14 (Wells et al, 1994).
We demonstrated that morphology in combination with blast cell reduction on day 15 could be used as a prognostic parameter identifying a standard-risk group. This group included the known favourable cytogenetic subtypes, but, in addition, 38% more patients with normal or other karyotypes with the same good prognosis (Table VII). However, in children with HR criteria and favourable karyotypes prognosis is favourable, indicating that these children can be moved to the SR group (Fig 3).
The relationship between a high leukaemic cell mass and prognosis is well known, particularly in acute lymphoblastic leukaemia, and may be explained by the occurrence of mutant drug-resistant leukaemic cells (Russell, 1997). Hyperleucocytosis was also found to be an independent risk factor in our studies, but was not included in our risk stratification for SR patients for two reasons: (1) Only a few patients with hyperleucocytosis and favourable morphology qualified for the SR group by virtue of the association between WBC and blast cell reduction on day 15. (2) Hyperleucocytosis is mainly a factor which further increases the risk in HR patients. This very-high-risk group urgently requires new treatment strategies, including better supportive care during the first days after admission (Ablin, 1984; Creutzig et al, 1987a, b).
Age was not an independent factor due to the high association of young age and morphological subtypes of M5 or M7, hyperleucocytosis and unfavourable karyotypes (Vormoor et al, 1992).
Regarding immunophenotyping, the results of study AML-BFM 87 (Creutzig et al, 1995) confirmed the reports of the Children's Cancer Study Group (Smith et al, 1992), and the Pediatric Oncology Group (Kuerbitz et al, 1992), in that immunological markers were not of prognostic significance.
The risk definition used in our trials allowed a risk-adapted treatment for SR patients, reducing the risk of treatment mortality and severe late toxicity. In study BFM 93, these patients received a lower cumulative anthracycline dose and no allogeneic BMT in first CR. For HR patients, more experimental drugs and therapy are justified during induction and consolidation, to increase the rate and quality of remission, and, in addition, allogeneic BMT in first remission from matched related donor is recommended. The definition of a third very-high-risk group due to hyperleucocytosis is possible and may be useful as soon as new effective treatment options become available.
Principal investigators of studies AML-BFM 83 and 87 in Germany: R. Mertens, Kinderklinik RWTH, Aachen; A. Gnekow, I. Kinderklinik des Klinikums, Augsburg; R. Dickerhoff, Johanniter Kinderklinik, St. Augustin; G.F. Wündisch, Universitäts-Kinderklinik, Bayreuth; G. Henze, Kinderklinik der Freien Universität, Berlin; U. Bode, Universitäts-Kinderklinik, Bonn; H.-J. Spaar/Th. Lieber Prof.-Hess-Kinderklinik, Bremen; W. Eberl, Städtische Kinderklinik, Braunschweig; W. Andler/I. Meyer, Vestische Kinderklinik, Datteln; U. Göbel/H. Jürgens, Universitäts-Kinderklinik, Düsseldorf; J.D. Beck, Universitäts-Kinderklinik, Erlangen; W. Havers/B. Stollmann-Gibbels Universitäts-Kinderklinik, Essen; B. Kornhuber, Universitäts-Kinderklinik, Frankfurt; Ch. Niemeyer, Universitäts-Kinderklinik, Freiburg; M. Lakomek/A. Pekrun, Universitäts-Kinderklinik, Göttingen; F. Lampert/R. Blütters-Sawatzki, Universitäts-Kinderklinik, Gieβen; K. Winkler, Universitäts-Kinderklinik, Hamburg; H. Riehm, P. Weinel, Kinderklinik der Medizinischen Hochschule, Hannover; R. Ludwig, B. Selle, Universitäts-Kinderklinik, Heidelberg; N Graf/M. Müller, Universitäts-Kinderklinik, Homburg/Saar; G. Nessler, Städtische Kinderklinik, Karlsruhe; Th. Wehinger, Städt. Kinderklinik, Kassel; M. Rister, Kinderklinik Kemperhof, Koblenz; F. Berthold, Universitäts-Kinderklinik, Köln; W. Sternschulte, Städtisches Kinderkrankenhaus, Köln; R. Schneppenheim, Universitäts-Kinderklinik, Kiel; P. Bucsky, Universitäts-Kinderklinik, Lübeck; O. Sauer, Städt. Kinderklinik, Mannheim; P. Gutjahr, Universitäts-Kinderklinik, Mainz; R. Eschenbach, Universitäts-Kinderklinik, Marburg; W. Müller, Kinderklinik d. Krankenhaus Neuwerk, Mönchengladbach; R.J. Haas, v. Haunersches Kinderspital, München; St. Müller-Weihrich, Kinderklinik d. Technischen Universität, München-Schwabing; Ch. Bender-Götze/R. Köglmeier, Universitäts-Kinderpoliklinik, München; P. Klose, Städtische Kinderklinik, München-Harlaching; G. Schellong, Universitäts-Kinderklinik, Münster; A. Jobke, Cnopfsche Kinderklinik, Nürnberg; U. Schwarzer, Städtische Kinderklinik, Nürnberg; J. Treuner, Olgahospital, Stuttgart; D. Niethammer/H. Scheel-Walter, Universitäts-Kinderklinik, Tübingen; W. Behnisch, Universitäts-Kinderklinik, Ulm; J. Kühl, Universitäts-Kinderklinik, Würzburg.
The coordinators of studies AML-BFM 83 and 87 were G. Schellong, J. Ritter, and U. Creutzig, Universitäts-Kinderklinik, Münster.
We thank P. Stappert, E. Kurzknabe and J. Meltzer for excellent technical assistance and Christa Lausch for her valuable assistance in the management of the AML studies. This work was supported by Deutsche Krebshilfe Germany.