Impact of age on outcome of pediatric acute myeloid leukemia

A report from 2 institutions

Authors


  • Presented in part at the 46th Annual Meeting of the American Society of Hematology, December 5, 2004, San Diego, CA.

Abstract

BACKGROUND

The prognostic significance of age among pediatric patients with acute myeloid leukemia (AML) was investigated.

METHODS

The authors reviewed the outcome of 424 patients who were ≤21 years of age at the time of diagnosis of AML (excluding acute promyelocytic leukemia) between 1983 and 2002 at St. Jude Children's Research Hospital (n = 288) or the M. D. Anderson Cancer Center (n = 136). Two treatment eras (1983-1989 and 1990-2002) were examined because of the greater intensity of treatment during the recent era.

RESULTS

After controlling for the effects of cytogenetics, white blood cell (WBC) count, French–American–British (FAB) subtype, and treatment era, they observed that age and treatment era interacted significantly in relation to event-free survival (EFS) (P = .033). Patients 10 years of age or older were at greater risk of an adverse event than younger patients in the recent era (hazard ratio = 1.8; 95% confidence interval [CI]: 1.3-2.6; P = .005) but not in the early era. The rate of adverse events (death or recurrence) increased significantly with each year of age in the recent era (4.3%/year; 95% CI: 1.9-6.8%; P = .001) but not in the early era. The rate of death increased significantly with each year of age in both eras (4.4%/year; 95% CI: 2.3-6.5%; P<.001). EFS and survival showed no association with study site, and the effects of age were similar at the 2 sites.

CONCLUSIONS

These results suggest that age is an independent prognostic factor in childhood AML and that children younger than 10 years benefit more than older children from newer intensive therapies. Cancer 2006. © 2006 American Cancer Society.

The outcome of pediatric acute myeloid leukemia (AML) has improved over the past 3 decades, and the cure rate has reached 50% with current therapy.1–5 The outcome of adults with AML is poorer,6 especially among elderly patients, whose disease is often associated with adverse prognostic factors.7 The impact of age on prognosis in pediatric patients with AML is not well established. Younger children treated on the Medical Research Council AML 10 and 12 trials had a lower rate of recurrence and higher overall survival (OS) and event-free survival (EFS) estimates than their older counterparts.8 Although a Japanese study showed a trend toward decreasing EFS with increasing age (43.1% for ages 1 to 9 years, 34.3% for ages 10 to 15 years, and 32% for ages 15 to 19 years),9 the difference was not statistically significant; moreover, overall survival rates of these 3 age groups were similar.9

A recent study of the experience of St. Jude Children's Research Hospital indicated that children ≥10 years of age were at increased risk of death during induction therapy or postremission therapy for AML.10 In a study of autologous stem cell transplantation for childhood AML, transplant-related mortality, treatment failure, and overall mortality rates were higher among those > 10 years of age.11 In another study of allogeneic bone marrow transplantation in children with AML in first remission, the overall survival of children ge; 10 years of age was less than that of younger children because of the higher risk of severe graft-versus-host disease.12

In this study, we determined the impact of age on treatment outcome in pediatric patients with newly diagnosed AML treated at 2 referral centers. We also analyzed the effect of recent treatment advances during the past 10 years on prognosis in different age groups.

MATERIALS AND METHODS

Patients

We reviewed the records of all 424 patients who were ≤21 years of age at diagnosis of AML (excluding acute promyelocytic leukemia) between 1983 and 2002 at St. Jude Children's Research Hospital (SJCRH) and at the M. D. Anderson Cancer Center (MDACC). The 288 St. Jude patients were enrolled on the frontline AML83,13 AML87,14 AML91,15 and AML9716 protocols, and the 136 patients at MDACC were enrolled on multiple institutional protocols17, 18 and on Children's Cancer Group (CCG) protocols 2891 and 2961.19 All chemotherapy administered, at both institutions, was dosed based on body surface area, which was calculated based on actual not ideal body weight for all patients. Patients were divided according to 2 treatment eras. Patients in the early era (n = 266) consisted of those who began treatment between 1983 and 1989 at MDACC and patients who were enrolled on the AML83 or AML87 protocols between 1983 and 1990 at St. Jude. Patients in the recent era (n = 158) began treatment at MDACC during the period 1990 through 2002 or were enrolled on the AML91 or AML97 protocols at St. Jude. The 2 eras were defined slightly differently at the 2 centers in order to encompass all patients treated in the individual protocols. Higher-intensity treatment protocols were used during the recent era at both institutions. At SJCRH, higher intensity treatment consisted of administration of cladribine (2-CDA) and more frequent utilization of stem cell transplantation. At MDACC, intensive timing of chemotherapeutic agents was utilized on the CCG 2961 protocol,19 as well as administration of idarubicin on CCG and institutional protocols.17, 18 Of the 93 AML patients 10 or more years of age who were treated at MDACC, 27 (29%) were treated on the pediatric service and 66 (81%) on the adult service.

Age was grouped (<10 or ≥10 years) based on using the rounded median age (9.8 years) as a cut point. It should be noted that this cut point is consistent with other studies.9–12

All patients were assigned to 1 of 3 cytogenetic risk groups: favorable [inv(16) or t(16;16); t(9;11); t(8;21);20 Down syndrome21], intermediate [normal cytogenetics or not classifiable as favorable or adverse], and adverse [-7; -5; del (5q); abnormal 3q; complex karyotype].20 It is noteworthy that, whereas t(9;11) is associated with favorable treatment outcome in the context of therapy used at SJCRH22 and elsewhere,5 its prognostic significance has been interpreted differently in the context of other chemotherapy regimens.23

The median (range) follow-up for survivors was 8.5 (2.1-19.6) years for St. Jude patients and 5.4 (0.2-16.9) years for MDACC patients. The combined median follow-up for survivors was 7.7 years.

Statistical Analysis

Patient characteristics were compared according to study site, age (<10 years vs. ≥10 years), and service by using the chi-square or exact chi-square test for categorical variables,24 and the Wilcoxon rank sum test for continuous variables.25 OS and EFS were estimated by using the method of Kaplan and Meier, with standard errors calculated by the method of Peto and Pike.26 EFS was defined as the time between protocol enrollment and recurrence, death due to any cause, or the last follow-up visit. Survival distributions were compared by using Mantel–Haenszel tests,27 with stratification based on treatment era and/or site depending on the comparison.

Survival estimates were compared according to age group by using Cox proportional hazards models28 adjusted for gender, race (white, black, or Hispanic/other), cytogenetic risk group (favorable, intermediate, adverse), white blood cell (WBC) count at diagnosis (≤50 vs. >50 × 109/L), French–American–British (FAB) subtype (M7 vs. other), study site, and era. Age and WBC count were also investigated as continuous variables in separate models. Stepwise variable selection, performed using SAS Release 9.1 software (SAS Institute, Cary, NC), was used to reduce the number of variables in the model; variables were retained and entered at the P = .15 level of significance. Interactions of interest were explored by using contrasts, and bone marrow transplantation was investigated as a time-varying covariate after stepwise variable selection.

The cumulative incidence of bone marrow recurrence and/or refractory disease was estimated,29 with death in complete remission, death during induction, and second malignancy treated as competing events. Here, follow-up was censored at transplant to be consistent with the study by Rubnitz et al.10 The Gray test30 was used to compare the cumulative incidence estimates across age groups.

RESULTS

Patient Characteristics

Patients at the 2 centers differed in the distribution of age, race, and cytogenetic risk group (Table 1). MDACC patients were significantly older, less likely to be black, more likely to be Hispanic, and more likely to have adverse cytogenetic features than were St. Jude patients. MDACC patients were slightly less likely to have the M7 FAB subtype. Bone marrow transplantation was performed more frequently at St. Jude (33.3% of patients) than at MDACC (4.4%).

Table 1. Comparison of Patient Characteristics at Diagnosis at the 2 Study Sites
CharacteristicSt. Jude (n = 288) No.%MDACC (n = 136) No.%P
  • SJCRH: St. Jude Children's Research Hospital; MDACC: M. D. Anderson Cancer Center; WBC: white blood cell; FAB: French-American-British.

  • *

    Data missing for 1 patient in the SJCRH group.

Age (yrs)     
 <1017159.44331.6<.001
 ≥1011740.69368.4 
Sex     
 Male14550.47756.6.228
 Female14349.65943.4 
Race     
 White20470.85540.4<.001
 African American5218.11611.8 
 Hispanic3010.46044.1 
 Other20.753.7 
Cytogenetic features     
 Favorable9332.44130.8.018
 Intermediate18163.17657.1 
 Adverse134.51612.1 
WBC count ×109/L     
 ≤5020771.99267.7.373
 >508128.14432.4 
FAB subtype*     
 M072.410.7.077
 M143152719.9 
 M27927.64835.3 
 M44816.72417.7 
 M56121.21813.2 
 M610.410.7 
 M73512.285.9 
 Other/unclassified134.596.6 

The age groups (≥10 years vs. <10 years) differed significantly in the distribution of FAB morphology and cytogenetic features. Patients in the older age group were less likely to have M7 and M4/M5 FAB morphology but were more likely to have M1 and M2 morphology (Table 2). They were also more likely to have intermediate and adverse cytogenetic features than the younger age group (Table 2).

Table 2. WBC Count, FAB Subtype, and Cytogenetic Risk Group Presented According to Patient Age
CharacteristicAge groupP
<10 years (n = 214)≥10 years (n = 210)
N%N%
  • SJCRH: St. Jude Children's Research Hospital; WBC: white blood cell; FAB: French-American-British.

  • *

    Data missing for 1 patient in the SJCRH group.

WBC count × 109/L     
 ≤5015974.314066.7.085
 >505525.77033.3 
FAB subtype*     
 M052.431.4<.001
 M1209.45023.8 
 M24320.28440 
 M44018.83215.2 
 M55123.92813.3 
 M610.510.5 
 M74119.321 
 Other/unclassified125.6104.8
Cytogenetics     
 Favorable7635.75828.0183
 Intermediate12558.713263.8 
 Adverse125.6178.2 

Effect of Age and Treatment Era on Outcome

Univariate analysis.

Estimates of EFS (P = .003) and OS (P = .002) differed significantly in a site- and era-stratified comparison. Older patients (≥10 years) had lower 5-year EFS (28.3±3.8% vs. 40.3±4.1%) and OS (34.8±4.0% vs. 49.4±4.0%) estimates than did younger patients (Fig. 1). This difference in outcome could be explained, in part, by the higher incidence of bone marrow recurrence and/or refractory disease in patients ≥10 years of age. The 5-year cumulative incidence of bone marrow recurrence and/or refractory disease was significantly higher in patients age ≥10 years than in the younger age group (69.7% ± 3.7% vs. 57.2% ± 4.2%; P = .006). EFS showed a significant age group difference in the recent era (P = .006) but not in the early era (P = .593) and differed significantly between eras for younger patients (P = .012) but not for older patients (P = .813). The 5-year EFS estimate was 25.6% ± 4.7% for the early era vs. 50.4% ± 5.8% for the recent era among patients <10 years old and 25.4% ± 5.0% in the early era vs. 29.7% ± 5.4% in the recent era for those ≥10 years old (Fig. 2). Similarly, OS differed significantly according to age group in the recent era (P = .006) but not in the early era (P = .116). These results indicate that outcome has improved in the recent era for patients <10 years old but not for patients ≥10 years old.

Figure 1.

Kaplan–Meier estimate of the probability of (a) survival and (b) event-free survival in 424 pediatric patients with acute myeloid leukemia (AML), according to age at diagnosis (<10 years vs. ≥10 years).

Figure 2.

Kaplan–Meier estimate of the probability of event-free survival in 424 pediatric patients with acute myeloid leukemia (AML), according to age at diagnosis (<10 years vs. ≥10 years) and treatment era (early vs. recent).

Multivariate regression analysis.

A multivariate Cox regression analysis showed that age remained a significant predictor of both OS and EFS after adjustment for other significant covariates (Table 3). Patients ≥10 years of age were at greater risk of death than younger patients after adjustment for cytogenetic risk group, WBC count, M7 FAB subgroup, and treatment era (hazard ratio = 1.63; 95% confidence interval [CI]: 1.23-2.16; P = .001). After adjustment for covariates, no significant interaction was seen between age group and era for OS (P = .712). For EFS, there was a significant interaction between age group and treatment era (P = .033); older patients were at significantly greater risk of an adverse event than younger patients in the recent era (hazard ratio = 1.83, 95% CI: 1.30-2.57; P = .005) but not in the early era. Younger patients were more likely to experience an adverse event in the early than in the recent treatment era (hazard ratio = 1.65, 95% CI: 1.15-2.35; P = .006), but for the older patients no improvement was observed in the recent era. Moreover, results of the multivariate analysis suggested that patients with intermediate and adverse cytogenetics, WBC count >50 × 109/L at diagnosis, and M7 FAB subtype were at increased risk of an adverse event.

Table 3. Multivariate Analysis of Overall and Event-Free Survival (n = 420)
Variable*Hazard ratio95% confidence limitsP
  • WBC: white blood cell; FAB: French-American-British.

  • *

    Reference groups are age <10 for age ≥10, favorable cytogenetic group for intermediate and adverse cytogenetic groups, all other FAB categories for FAB M7, and recent era for early era, respectively.

  • Because there was a significant interaction between age group and era in relation to event-free survival, hazard ratios are presented for age group in each era and for era in each age group. The age group effect gives the hazard ratio for patients 10 and older relative to those less than 10 years. The era effect gives the hazard ratio for patients diagnosed in the early era relative to those diagnosed in the recent era.

Overall survival    
 Age ≥10 years at diagnosis1.6301.2302.160.001
 Intermediate cytogenetic group2.1821.5932.990<.001
 Adverse cytogenetic group3.1711.8925.313<.001
 WBC >50 × 109/L at diagnosis1.4031.0671.844.015
 FAB M72.7121.7654.167<.001
 Early era (before 1990/91)1.4831.1451.919.003
Event-free survival model    
 Intermediate cytogenetic group1.9501.4632.600<.001
 Adverse cytogenetic group3.4042.1115.487<.001
 WBC >50 × 109/L at diagnosis1.5551.2032.011.001
 FAB M72.0051.3203.047.001
Age group × era   .033
 Age group effect in the early era1.0780.7421.304.694
 Age group effect in the recent era1.8281.3002.571.005
 Era effect for age ≥10 group0.9710.6951.356.864
 Era effect for age <10 group1.6471.1532.353.006

When age was modeled as a continuous variable, the risk of death increased 4.4% with every additional year of age at diagnosis (95% CI: 2.3-6.5%; P<.001) after adjustment for cytogenetic risk group, WBC count, and M7 FAB subtype. In the recent era, the risk of an adverse event increased 4.3% for every year of increase in age (95% CI: 1.9-6.8%; P = .001), but it did not increase significantly with age in the early era (results not shown).

The 10 patients with Down syndrome were all < 5 years of age. Given that Down syndrome is a favorable risk factor that could potentially bias the difference between the 2 age groups, the multivariate analyses were performed without these patients, and the results were unchanged qualitatively and similar quantitatively (results not shown).

Effect of Bone Marrow Transplantation

Of the 424 patients, 102 (24%) received bone marrow transplants; 10 of 158 patients (6.3%) received transplants in the early era, with 3 being autologous, and 92 of 266 (34.6%) in the recent era, with 47 being autologous. Hence, it is difficult to ascertain whether the apparent era-related benefit reflects the increased use of transplantation or whether the apparent transplant benefit reflects the effect of era. An analysis of only the MDACC patients, of whom only 6 underwent transplantation, found similar results as in the combined cohort; therefore, improvement observed over time may reflect more than the impact of transplantation. Older patients were less likely to have received transplants (16.7% of patients ≥10 years old vs. 31.3% of patients <10 years old). In a multivariate model, after adjustment for transplantation, age group was still a significant predictor of OS (results not shown) and of EFS (Table 4).

Table 4. Multivariate Analysis of Factors Affecting Event-Free Survival in the Recent Era (after 1990/91)
Variable*Hazard ratio95% confidence limitsP
  • WBC: white blood cell; FAB: French-American-British.

  • *

    Reference groups are age <10 years, favorable cytogenetic risk group, WBC ≤50 × 109/L, all other FAB categories, and no bone marrow transplant, respectively.

Age ≥10 at diagnosis1.8971.3022.762.001
Intermediate cytogenetic group2.0731.3963.077<.001
Adverse cytogenetic group4.0612.3237.098<.001
WBC >50 × 109/L at diagnosis1.8331.2802.626.001
FAB M72.7301.5944.674<.001
Bone marrow transplant0.9690.6571.429.873

Effect of Age on Outcome According to Treatment Site and Service

Treatment site (MDACC vs. SJCRH) did not significantly influence EFS (P = .073) or OS (P = .40) after adjusting for era, and the relation of age to EFS (Fig. 3) and OS (data not shown) was similar at the 2 sites. When site was investigated separately for each era, it showed no effect on outcome in either era.

Figure 3.

Kaplan–Meier estimate of the probability of event-free survival of 424 pediatric patients with acute myeloid leukemia (AML), according to age at diagnosis (<10 years vs. ≥10 years) and treatment site (St. Jude Children's Research Hospital [SJCRH]; M. D. Anderson Cancer Center [MDACC]).

For patients ≥10 years of age (n = 93) treated at MDACC, the 5-year EFS estimates for patients treated by the pediatric service (n = 27) and the adult service (n = 66) were similar (18.9 ± 8.5% and 23.3 ± 6.8%, P = .892). Likewise, the OS estimates for pediatric service and adult service patients were not found to differ significantly (45.9 ± 11.9% vs. 28.4 ± 6.9%, P = .181). It should be noted that age significantly differed between the 2 services (P<.001). The median (range) age was 13 (10.1-16.6) years for the pediatric service and 20.1 (15.5-21.8) for the adult service.

DISCUSSION

This retrospective study of 424 children and adolescents with newly diagnosed AML treated at 2 different institutions over 20 years indicates that with recent improvement in treatment outcome, age at diagnosis has become an independent prognostic factor. Patients < 10 years old have significantly better outcome than those >10 years of age, regardless of treatment institution. Studies in adults have reported the adverse effect of age > 60 years,7, 31 which is attributed to both disease biology and host factors. Elderly patients have a high incidence of intermediate and unfavorable cytogenetics, antecedent hematological disorders such as myelodysplasia, multidrug resistance gene expression, and FLT3 internal tandem duplication.7, 31–33 In addition, associated comorbidities and advanced age preclude the ability to deliver optimal intensive therapy.31, 34 In our patient series, the proportion of patients with intermediate or adverse cytogenetics was higher among those 10 or more years of age (71%) than among those in the younger age group (64.3%; P = .15), a finding that may have contributed to the poorer prognosis in this age group. In the Medical Research Council AML 10 and 12 trials, the distribution of FAB subtypes was reported to vary with age.8 As in our study, FAB subtypes M5 and M7 were more common in early childhood, whereas older children were more likely to have FAB subtypes M1 and M2. In a multivariate analysis, OS and EFS estimates were higher in younger children due to a lower rate of bone marrow recurrence. In our series, the cumulative incidence of bone marrow recurrence and/or refractory disease was significantly higher in patients 10 years or more of age than in the younger age group.

Although the status of FLT3 internal tandem duplication (FLT3-ITD) was not determined in our patients, FLT3-ITD was recently reported to be more prevalent in older children with AML35; 1% in those 0-1 years, 6% in those 1-5 years, 14% in those 5-10 years, and 19% in those 10-20 years of age. Because this genetic abnormality was associated with poor outcome in pediatric AML,36 its higher prevalence in older children could have contributed to their worse treatment outcome in our study.

Another factor that could contribute to the worse outcome in the older age group is their increased risk of death during both induction and postremission therapy, as noted in a recent study of 260 St. Jude patients,10 95% of whom are included in this report.

In our earlier study,10 the 2-year estimates of death were 0% for infants, 2.3% ± 1.3% for children 1-9 years old, and 12.4% ± 3.6% for patients ≥10 years old. Similarly, in this series the 2-year cumulative incidence of death unrelated to leukemia was 0% for infants, 3.8% ± 1.5% in children 1-9 years old, and 8.3% ± 2.0% in the older age group.

Unlike in AML, age at diagnosis determines risk category and therapy in pediatric acute lymphoblastic leukemia (ALL). Children 1-9 years of age have a better outcome than those who are 10 or older, despite more intensive therapy in the latter group.37 In addition, 3 reports have shown a better treatment outcome in adolescents and young adults treated on pediatric rather than adult ALL protocols.38–40 Differences in treatment regimens, protocol compliance by the patients and physicians, adherence to treatment guidelines, and other factors41–44 could have contributed to this discrepancy in treatment outcome. In this series, despite the small numbers of patients, we found no significant difference in the EFS and OS estimates of the 27 AML patients treated by the pediatric service and the 66 patients treated by the adult service in 1 center. Because patients with AML received most if not all therapy intravenously in the hospital setting, noncompliance with medications was not an issue. Moreover, regardless of age, all patients were treated by an expert team in an academic setting.

Race has been shown to be a prognostic factor in some studies.45, 46 Therefore, OS and EFS were also investigated according to race, because the racial distribution differed significantly between the 2 institutions. We found no significant differences in outcome according to race at either institution (results not shown), and race did not enter into any of the multivariate models.

An interesting finding of this study is that patients with AML who were <10 years of age had a significantly better outcome in the recent treatment era, whereas their older counterparts did not. This benefit was related to intensification of chemotherapy rather than to stem cell transplantation, although the effect of transplantation could not be evaluated fully in this study because of the small number of patients who underwent transplantation, especially in the MDACC group.

Our results suggest that age is an independent prognostic factor in childhood AML and that children <10 years of age benefit more from newer intensive treatment than do their older counterparts. Older children do not appear to benefit from intensified treatment regimens introduced over the past decade, including autologous and allogeneic bone marrow transplantation, which are associated with more treatment-related toxicity and mortality in older children.11, 12 It would be desirable to discover whether pharmacokinetic differences account for the higher rate of treatment-related death among children >10 years of age. Ongoing clinical trials will determine whether new therapeutic approaches, including agents with novel targets and reduced-intensity conditioning for allogeneic bone marrow transplantation, will improve the outcome of older children with AML.

Acknowledgements

The authors thank Sharon Naron, M.P.A., E.L.S., for expert editorial assistance.

Ancillary