Anemia and survival in childhood acute lymphoblastic leukemia

Abstract

Background Several studies have demonstrated that patients with childhood acute lymphoblastic leukemia presenting with mild anemia at diagnosis have an increased risk of poor outcome compared to patients with more severe anemia. However, it has not been reported whether there is any correlation between degree of anemia and leukemia subtype.Design and Methods In a cohort of 1162 patients with childhood acute lymphoblastic leukemia we analyzed whether there was a correlation between degree of anemia and leukemia subtype. We also studied the association between degree of anemia and event-free survival within the subtypes.Results Hemoglobin levels at diagnosis were distributed in a non-random pattern. The degree of anemia was significantly different for three distinct groups of patients compared to the remaining patients (mean hemoglobin; T-cell leukemia: 106 g/L versus 76 g/L (precursor B-cell acute lymphoblastic leukemia); within precursor B-cell ALL: TEL-AML1 positive: 68 g/L versus 79 g/L; BCR-ABL positive: 93 g/L versus 76 g/L; each p<0.05). Furthermore, in contrast to the entire study group, patients with T-cell leukemia, TEL-AML1⁺, and BCR-ABL⁺ precursor B-cell leukemia had a more favorable prognosis if presenting with a higher hemoglobin level (≥80 g/L).Conclusions These observations indicate that the formerly reported direct correlation between severity of anemia and survival in childhood acute lymphoblastic leukemia mainly reflects differences in the degree of anemia between distinct biological subgroups with different treatment outcomes. On the other hand, the inverse relationship between severity of anemia and survival found within specific subgroups suggests that very low hemoglobin levels at diagnosis are associated with more advanced disease in these subgroups.

Introduction

Anemia is common in patients with newly diagnosed childhood acute lymphoblastic leukemia (ALL). Several studies have demonstrated a correlation between degree of anemia and survival.1–4 Evidence suggests that patients presenting with higher hemoglobin (Hb) levels at diagnosis have an increased risk of poor outcome compared to patients with lower Hb levels. However, in multivariate models, Hb was not identified as an independent risk factor.3,4 Despite the fact that Hb is not applicable for risk stratification in clinical trials, the relationship between anemia and prognosis might hold some important biological information that warrants further investigation.

The pathomechanism of anemia in childhood ALL is not completely understood. Anemia in children with cancer is associated with decreased erythropoietic activity but not with inadequate erythropoietin production, leading to the assumption that anemia in patients with leukemia mainly results from suppression of normal hematopoiesis in the bone marrow by infiltrating blasts.5,6

To further understand the pathophysiology of anemia in childhood ALL, it is important to consider the heterogeneity of diseases in these patients, including immunological subgroups (precursor B-cell ALL versus T-cell leukemia) and cytogenetic subgroups (e.g. TEL-AML1, E2A-PBX1, BCR-ABL, hyperdiploid karyotype with >50 chromosomes, and MLL gene rearrangements).

The differences in clinical features at presentation between patients with T-cell leukemia and precursor B-cell ALL are well recognized and include higher Hb levels in patients with T-cell leukemia.7–9 However, the prognostic influence of the initial Hb level in T-cell leukemia is controversial. In some studies, a higher Hb level was associated with a worse prognosis.10,11 In other series, there was no association, or even an opposite association.12,13 The Hb level at diagnosis is apparently not suitable for application in risk stratification strategies in current clinical trials.14

Precursor B-cell ALL is a heterogeneous disease comprising several distinct biological subtypes that are associated with different outcomes. Patients with t(1;19) leading to a E2A-PBX1 gene fusion, hyperdiploidy involving more than 50 chromosomes, or TEL-AML1 gene fusion have a favorable outcome, whereas those with t(4;11) leading to a MLL-AF4 gene fusion and a t(9;22) leading to a BCR-ABL gene fusion have a dismal prognosis.15 Despite the accurate analysis of the incidence and clinical relevance of these subtypes, it has not been reported whether there is any relationship between cytogenetic subtypes and degree of anemia.16–21

In the present study, we evaluated whether there is a correlation between degree of anemia and leukemia subtype. Furthermore, we studied the association between degree of anemia and event-free survival (EFS) within these subtypes of leukemia.

Design and Methods

The present study utilized initial diagnostic characteristics of patients with ALL enrolled in the multicenter ALL-BFM 95 trial of the Berlin-Frankfurt-Munster (BFM) study group between April 1995 and June 2000 who underwent screening for fusion genes (TEL-AML1, BCR-ABL, MLL-AF4).22 Differences in the distribution of parameters between subgroups of patients were examined using the χ or Fisher’s exact test. Survival probabilities were calculated using the Kaplan and Meier method with differences compared by the log-rank test; standard error were calculated according to Greenwood.23–25 The probability of EFS (pEFS) was calculated from the date of diagnosis to the first event (death from any cause, tumor progression/relapse or non-response, or second malignancy) or to the date of last follow-up. The prognostic impact of anemia on treatment outcome was analyzed together with other known prognostic factors using Cox regression analysis.26 Statistical analyses were performed using the SAS program (SAS-PC, version 9.1, Cary, NC: SAS Institute Inc.). The patients’ follow-up data were updated as of October 2005.

Results

Initially, we analyzed the Hb levels and the relation between anemia and survival in the entire group of patients. Based on 1162 peripheral blood samples, the mean Hb level at diagnosis was 80 g/L (standard deviation 27 g/L) (Table 1). Patients who presented with higher Hb levels at diagnosis (Hb ≥80 g/L) had an increased risk of poor outcome compared to patients with lower Hb levels (Hb <80 g/L) (pEFS: 0.76 versus 0.8, log-rank p=0.05).

Table 1.Hemoglobin levels (in g/L) according to the immunological and cytogenetic subtypes of childhood acute lymphoblastic leukemia.

To characterize the distribution of anemia in the heterogeneous group of patients with childhood ALL, the Hb levels were analyzed according to immunophenotypic and cytogenetic subtypes of ALL (Table 1). Patients with T-cell leukemia (n=138) presented with significantly higher Hb levels at diagnosis than patients with precursor B-cell ALL (n=981) (mean Hb level, 106 g/L versus 76 g/L; p<0.001). Within the group of patients with T-cell leukemia, an inverse relation between Hb level and hyperleukocytosis (leukocytes ≥100×10/L) was found (Table 2A). There was no association between degree of anemia and age among patients with T-cell leukemia (Table 2B).

Within the heterogeneous group of precursor B-cell ALL, the lowest mean Hb level was associated with TEL-AML1 leukemia (n=241) (mean Hb level 68 g/L). The mean Hb level was 75 g/L in MLL-AF4 leukemia (n=11), 76 g/L in hyperdiploid karyotype leukemia (n=157), and 93 g/L in BCR-ABL leukemia (n=23). Considering only precursor B-cell ALL, the observed Hb levels were significantly different (p<0.05) for both the TEL-AML1 and BCR-ABL subgroups compared to the remaining precursor B-cell ALL group (Table 1). Again, the association between anemia and hyperleukocytosis, and the influence of age, was evaluated for these two subgroups. As for T-cell leukemia, an inverse relation was found between Hb level and hyperleukocytosis (leukocytes ≥100×10/L) which was, however, only statistically significant in the TEL-AML1 subgroup (Table 2A). Furthermore, younger age (<10 years) was associated with lower Hb levels at diagnosis in both groups (Table 2B).

Patients enrolled in the ALL-BFM 95 trial did not undergo screening for the E2A-PBX1 gene fusion. Therefore, anemia could not be analyzed separately in this subgroup of patients.

The second aim of this retrospective study was to evaluate whether there is an association between Hb level and prognosis within the subgroups. Due to a limited number of patients in the MLL-AF4 group, this evaluation could only be performed for T-cell leukemia, TEL-AML1, BCR-ABL, and hyperdiploid karyotype precursor B-cell ALL.

Figure 1.Kaplan-Meier estimate of event-free survival of 138 children with T-cell leukemia according to the hemoglobin level at diagnosis. The cut-off hemoglobin level of 80 g/L is the mean value of the entire study group.

Among children with T-cell leukemia, the Hb level at diagnosis was significantly associated with EFS. Patients presenting with lower Hb levels at diagnosis had an increased risk of poor outcome compared to patients with higher Hb level [(pEFS: 0.77 (Hb ≥80 g/L, n=114) versus 0.54 (Hb <80 g/L, n=24), p=0.03, and 0.82 (Hb ≥110 g/L, n=67) versus 0.65 (Hb <110 g/L, n=71), p=0.03)] (Figures 1 and 2). There was a trend for a similar difference in the TEL-AML1 subgroup. Again, an advantage was observed for patients with higher Hb level as compared to those with lower Hb levels [(pEFS: 0.97 (Hb ≥80 g/L, n=66) versus 0.89 (Hb <80 g/L, n=175), log-rank p=0.12, 6-year pEFS p=0.01)]. In addition, there was a significant association between anemia and EFS within the BCR-ABL subgroup. Treatment failed in all patients who were diagnosed with BCR-ABL rearrangement in combination with lower Hb level at diagnosis [(pEFS: 0.35 (Hb ≥80 g/L, n=17) versus 0 (Hb <80 g/L, n=8), p=0.02)]. No substantial association was found between Hb level and prognosis for hyperdiploid karyotype leukemias [(pEFS: 0.85 (Hb ≥80 g/L, n=74) versus 0.88 (Hb <80 g/L, n=86), log-rank p=0.52, 6-year pEFS p=0.51)].

Table 2.Hemoglobin levels (in g/L) according to initial leukocyte count (A) and age (B).

Figure 2.Kaplan-Meier estimate of event-free survival of 138 children with T-cell leukemia according to hemoglobin level at diagnosis. The cut-off hemoglobin level of 110 g/L is the mean value within the group of patients with T-cell leukemias.

To further evaluate the prognostic impact of Hb in T-lineage and TEL-AML1 ALL, Cox models were applied including factors such as Hb level (≥80 g/L) and leukocyte count (≥100×10/L) at diagnosis, age (≥10 years), and prednisone response (the presence of ≥1,000 blasts per microliter of peripheral blood at day 8 was defined as a prednisone poor response). Due to the limited number of patients, no Cox model was constructed for the BCR-ABL group. In T-cell leukemia, only a prednisone poor response was an independent prognostic factor (RR=5.74, 95%CI 2.62–11.44, p<0.001). All data including those for Hb, leukocytes, and age are presented in Table 3A. Applying the Cox model to the TEL-AML1 subgroup, no independent risk factor could be identified (Table 3B). Thus, when adjusted for other risk factors, Hb is not an independent risk factor.

Discussion

Anemia is a consistent hematologic finding at the diagnosis of childhood ALL. Here, we confirm the results of former studies in which a relationship between degree of anemia and outcome was observed. Severe anemia at diagnosis (Hb <80 g/L) was again associated with a better EFS than that in patients with only mild anemia (Hb ≥80g/L). However, the observed difference was only moderate and did not take into consideration the heterogeneity of childhood ALL.

Table 3.Cox regression analysis in T-cell leukemia (A) and TEL-AML1 positive precursor B acute lymphoblastic leukemia (B): Risk ratios and 95% confidence interval.

Until now, no studies evaluated a possible correlation between degree of anemia and leukemia subtype. Our study demonstrates that Hb levels at diagnosis are distributed in a non-random pattern in childhood ALL. As expected, patients with T-cell leukemia presented with significanty higher Hb levels at diagnosis than did patients with precursor B-cell ALL. Although the basis for this difference is not well understood, it is likely that biological regulation of bone marrow infiltration, cell clone proliferation, and circulating peripheral blasts substantially differentiate T-cell leukemia from precursor B-cell ALL. In addition to the difference in the Hb level between T-lineage ALL and precursor B-cell leukemias, we also found a statistically significant difference in Hb level between both BCR-ABL and TEL-AML1 ALL and precursor B-cell ALL without these fusions. Several studies have analyzed the clinical and biological features in ALL patients carrying the translocation t(12;21) leading to the TEL-AML1 gene fusion.27–38 None of those studies reported a low Hb level to be associated with TEL-AML1. One study addressed this issue, but failed to demonstrate a correlation between Hb and TEL-AML1, probably due to the limited number of patients analyzed (11/51 patients TEL-AML1).39 However, a possible association between lower Hb levels at diagnosis and the TEL-AML1 subtype were observed for the first time recently by a group from China.40 The analysis of 95 pediatric ALL patients including 20 TEL-AML1 cases revealed a significantly lower Hb level in the TEL-AML1 patients than in the other patients (61 g/L versus 76 g/L; p=0.003). Their findings have now been confirmed by our data. We add significant information having evaluated a much larger group of patients and having conducted a comprehensive subgroup analysis.

Our data strongly suggest that there is a correlation between leukemia subtype and the degree of anemia indicating distinct mechanisms in the formation of erythropoietic insufficiency. Lower Hb levels (Hb <80 g/L) were more often diagnosed in leukemia subtypes associated with a favorable outcome (TEL-AML1, hyperdiploid karyotype). In contrast, more aggressive leukemia subtypes (T-cell leukemia and BCR-ABL precursor B-cell leukemia) were associated with higher Hb levels (Hb ≥80 g/L).22 We suppose that the correlation between degree of anemia and survival is due to the correlation between degree of anemia and the distinct biological subgroups that are associated with different outcomes.

Evaluating the TEL-AML1 subgroup, anemia was more pronounced among younger patients (<10 years). There was a trend for a similar difference in the BCR-ABL subgroup. It is possible that lower baseline Hb levels in the younger population before developing leukemia might contribute to this finding; however, an additional impact of other biological factors cannot be excluded. Patients carrying the TEL-AML1 translocation are overrepresented in the younger age group (<10 years). We think it is rather unlikely that lower baseline Hb levels before developing leukemia exclusively contributes to the strong association of TEL-AML1 and severe anemia (Hb <80 g/L). This assumption is based upon the observation that the mean Hb level was lower in the TEL-AML1 subgroup than in the other subgroups in older patients (>10 years).

As regards the predictive power of Hb level in subgroups, the literature data on the prognostic influence of Hb in T-cell leukemia are controversial.10–13 Based on our data, higher Hb levels in this group of patients were clearly associated with better outcome compared to lower Hb levels. We observed a similar trend in the TEL-AML1 subgroup. Due to an unusual late relapse, there was no overall significance in this group using the log-rank test, however, EFS at 6 years was significantly different between the two groups. Despite a limited number of patients (n=25), estimation of EFS revealed striking differences within the subgroup of BCR-ABL precursor B-cell leukemia. Again, a low Hb level at diagnosis was an adverse prognostic marker. Despite the reported (univariate) association between anemia and EFS within the mentioned subgroups, the Hb level at diagnosis was not an independent risk factor and, therefore, of negligible value for risk assignment. The predictive power of treatment response was much stronger than that of the initial Hb level.

To summarize, initial degree of anemia in childhood ALL is non-randomly distributed. Moreover, in contrast to the group of patients considered as a whole, patients with T-cell leukemia, TEL-AML1, and BCR-ABL precursor B-cell leukemia have a more favorable prognosis if presenting with higher Hb levels. Given that it was observed in several series that a higher Hb level is associated with poorer outcome, it was assumed that this reflects conditions with a high proliferation rate of an aggressive leukemic cell clone (shorter history in fast disease).41,42 Based on our data, this hypothesis is not sustainable for (at least) the three ALL subtypes mentioned above. This is further supported by the observation that hyperleukocytosis was associated with lower Hb levels. We suggest that lower Hb levels or severe anemia at diagnosis might correspond to conditions of advanced disease. It is possible that patients presenting with higher Hb levels or mild anemia are detected at an early stage of disease, and are, therefore, more susceptible to chemotherapeutic interventions.

Footnotes

• Authorship and Disclosures OT: prepared the study, collected data, and wrote the manuscript; MS: designed the study, and contributed to the analysis and interpretation of the data and to writing the manuscript; GC, WDL: collected and analyzed data; SL: collected data and contributed to writing manuscript; BWS: analyzed and interpreted data and contributed to writing manuscript; MZ: analyzed data and performed the statistical analyses; MS: collected, analyzed and interpreted the data, designed the study and wrote manuscript; FKN: collected and interpreted data, designed the study and wrote the manuscript. The authors reported no potential conflicts of interest.

• Received April 1, 2008.
• Revision received July 18, 2008.
• Accepted July 21, 2008.