Introduction

Children with acute myeloid leukaemia (AML) often exhibit extramedullary infiltrates at diagnosis.1 The incidence of extramedullary leukaemia (EML) is reported to be between 20 and 40%.23 Sometimes, EML precedes the diagnosis of bone marrow involvement by weeks or months and poses a considerable therapeutic dilemma.4 The most common manifestations of EML are skin infiltrates, soft tissue or bone involvement, gingival infiltration and central nervous system (CNS) involvement. Testicular involvement is very rare in AML. The soft tissue tumours are also referred to as chloromas or in recent literature as granulocytic sarcomas or myeloblastomas.1

EML is more common in infants with AML,56789 and EML is also more common in the myelo-monocytic or monoblastic variants of AML (FAB M4 and M5), than in other morphological subtypes.310

There is considerable controversy about the prognostic significance of extramedullary involvement in AML. While some groups found that it conferred a poorer prognosis,1112 in other studies this does not seem to be the case.313

The chloromatous manifestation of AML seems to be more common in Africa and Asia, although it is not clear, whether genetic or socio-economic factors are responsible.1415 Blast cells from patients with myeloblastoma often exhibit the 8;21 chromosomal translocation.1617 Although the 8;21 translocation is generally associated with a relatively better prognosis, the presence of EML may compromise this prognostic advantage.1

Recently, Schwyzer et al18 reported from South Africa that AML with a myeloblastoma (chloroma) at diagnosis is associated with the 8;21 translocation and that these children have a better prognosis than those without myeloblastoma and t(8;21). Others reported an inferior prognosis for patients with myeloblastoma.1419

The controversies in the literature, concerning the incidence and prognostic significance of EML, prompted us to review retrospectively the population-based data of the DCLSG over the years 1972–1998, for the presence and prognostic significance of EML.

Patients and methods

Since 1972, children with leukaemia have been registered by the Dutch Childhood Leukaemia Study Group (DCLSG). This national registry is estimated to be 97% complete.20 Bone marrow and peripheral blood samples and smears of children suspected of having leukaemia are sent to the laboratory of the DCLSG for confirmation and classification of diagnosis and relapse. The diagnosis and FAB subtype was determined with May–Grünwald-Giemsa stained smears and additional stains, including Sudan-B-Black, myeloperoxidase, α-naphtyl acetate esterase with and without sodium fluoride. FAB classification was not available for the first 56 patients. The diagnosis of AML required the presence of >30% blast cells in the bone marrow and that leukaemic cells show morphological and cytochemical characteristics of myeloid or monocytic differentiation. Cytogenetic data were available for 170 of the 260 patients reported on a national level since 1989. Centralized phenotyping for AML patients was started in 1987.

Registration data and initial discharge summaries were reviewed for the presence of EML in all children with AML, aged 0–16 years, who were registered by the DCLSG from April 1972 to December 1998. EML was defined as a clinically obvious infiltrate in soft tissues, skin, muscles or bone, gingiva, cerebrospinal fluid or brain. CNS leukaemia was diagnosed on the basis of >5/μl blast cells in the CSF without blood contamination or as a leukaemic infiltration of the brain as evidenced by cranial CT scan. Complete remission was defined as <5% blast cells in the bone marrow, normal haematopoiesis and the absence of blast cells in the peripheral blood. Relapse was defined as a return of leukaemic blast cells in the bone marrow (>20%) and/or return of the EML.

Treatment protocols

A number of treatment protocols were used during the 26-year period of this retrospective study. Between 1972 and 1980 institutional protocols were used. From 1980 to 1982 patients received high-dose Ara-C, 6-thioguanine and daunorubicine induction, but no maintenance therapy. Those with HLA-identical siblings proceeded to bone marrow transplantation (BMT).21 From 1982 to 1985 induction treatment was individually ‘tailored’ on the basis of cells in S-phase. Patients entering remission went on to BMT or received the VAPA-10 protocol as maintenance.22 From 1985, treatment strategies were based on the BFM protocols, and the first national BFM-based AML study started in 1987.23 From 1997, the DCLSG has participated in the current UK MRC-12 AML Study.

Statistical methods

One-way ANOVA or two-sided t-test was used to test the differences between means of continuous variables. Pearson's chi-square test was applied to categorical variables with post-hoc comparisons. Event-free survival was calculated with the Kaplan–Meier method from the time of diagnosis to the first event (relapse or death of any cause) or censoring (date of last follow-up). Patients who did not achieve remission were considered as treatment failure on day 1.

The statistical difference between survival curves was calculated with the log-rank test. Cox regression was used for multivariate analysis. All statistical tests were carried out using SPSS for Windows (version 8). As the intensity of treatment is the most important determinant of prognosis, only patients diagnosed after 1 January 1985 were included in the analysis of event-free survival. Patients with an overlap (two forms of EML) were excluded from the subgroup analyses.

Results and discussion

There is conflicting information in the literature on the relative frequency of EML at diagnosis in AML. Some authors reported a frequency of 38–42%.212 This is rather more than the 25.1% found in our large, retrospective study. From April 1972 to December 1998, 477 consecutive patients (0–16 years old) were diagnosed with de novo AML and registered by the DCLSG. Of these patients, 120 had EML at presentation.

There is also little agreement in the literature over what constitutes an EML. While some authors include, for example, CNS leukaemia or hepatosplenomegaly, others do not.24 Among the patients reported here, 35 (7.3%) had soft tissue or bone involvement, 30 (6.3%) gingival infiltrates, 30 (6.3%) had CNS involvement and six (1.2%) had skin infiltrates. Nineteen patients (4%) had double sites of EML (Table 1). The localisation of the myeloblastoma was as follows; orbit 35, jaw four, mastoid one, breast one, upper extremity two, pancreas two, paraspinal one, buttocks two, lower extremities one. No isolated EMLs were found in the registry, all children with an EML had BM infiltration at diagnosis. The length of the pre-diagnostic history of soft tissue swelling varied between a few days and a few weeks.

Table 1  Number of patients with one or two forms of extramedullary leukaemia at diagnosis
Full size table

Characteristics in patients with or without EML are summarised in Table 2. The four types of EML found in this study correspond to the most frequently reported forms. Although the relative frequency of their occurrence varies between reported studies, these differences are not remarkable.122526

Table 2  Patient characteristics in the EML subgroups and in patients without EML
Full size table

When patients were split into two age groups, infants (age <1 year) with EML tended to fall into the myeloblastoma and CNS-positive group, had more skin infiltrates but there was only one infant with gingival infiltration. This corresponds to the higher mean age in the gingival infiltration group. Infants had a higher relative proportion of monoblastic leukaemia (FAB-M5) and presented less often with M1, M2 or M3 morphology. The event-free survival of infants was not significantly worse than that of children above 1 year of age.

In a univariate analysis of prognostic factors, initial WBC count and favourable cytogenetic findings (defined as: t(8;21) or t(15;17) or inv16 or trisomy 21) were significant factors (Table 3). In a stepwise multiple regression analysis only favourable cytogenetics remained significant (P = 0.033).

Table 3  Five-year estimatesa of event-free survival (EFS) in patients with acute myeloid leukaemia
Full size table

Myeloblastoma is often associated with FAB-M2 morphology and the 8;21 chromosomal translocation.116 While this translocation is generally associated with a better prognosis, the presence of EML (or more likely a molecular event behind it) might jeopardise this positive effect.1619 Genetic or socio-economic factors may further complicate this interaction. While children in Turkey and India had a poor outcome when presenting with orbital myeloblastoma,1415 their counterparts in South Africa seemed to have fared much better,18 and in most European or US studies and in a recent study from Saudi Arabia, the presence of myeloblastoma at diagnosis had no influence on survival.3121324

A possible explanation for the more infiltrative behaviour of leukaemic cells in EML could be an altered cell surface phenotype. Hurwitz et al27 reported in children with AML that the coexpression of CD19 and CD56 antigens is specific for the t(8;21) and suggested that the expression of the CD56 N-CAM neuronal adhesion molecule could explain the higher incidence of granulocytic sarcomas in patients with the above translocation. Seymour et al28 confirmed the coexpression of CD56 and CD19 in adult patients, but found a much smaller incidence of CD56 expression in the M2 cell type and no correlation with EML. Again, the same phenotype and morphology may mean something quite different in children than in adults.

Perhaps even more puzzling are our findings in children with gingival infiltrates. Gingival hypertrophy seems to be a special form of EML, where local microanatomic characteristics could play an additional role to that of the myeloid blast cells.29 Patients with gingival infiltrates in our series were older at diagnosis, had a high WBC count and a very high proportion of myelo-monocytic or monoblastic morphology (Table 2). Their EFS seems poorer than those, for example, with a myeloblastoma (Figure 1) but because of the small numbers one can draw no conclusions on this issue. Nevertheless, this observation calls attention to the need to prospectively investigate these patients, who may well require more intensive treatment in the future.

Figure 1
figure 1

 Kaplan–Meier estimate of event-free survival for patients with and without extra-medullary infiltrates at diagnosis. Logrank test for EFS, P = 0.85.

Full size image

Primary myeloblastoma, preceding the clinical diagnosis of AML should be treated with systemic chemotherapy, just as when they occur simultaneously.1 Permanent local control at sites of EML can be achieved without local therapy such as irradiation.24 The most important therapeutic question is whether patients with EML should receive more intensive treatment than others. Similarly to others, we found no significant difference in the EFS of patients with or without EML. In current studies, patients with EML at diagnosis are not allocated to a different or more intensive treatment arm. Analysis of the clinical and biological data from ongoing collaborative studies may throw more light on the pathomechanism and prognostic significance of EML in AML, particularly in patients with the t(8;21) karyotype.