Older age has historically been an adverse prognostic factor in pediatric acute myeloid leukemia (AML). To the authors' knowledge, the impact of age relative to that of other prognostic factors on the outcome of patients treated in recent trials is unknown.
Clinical outcome and causes of treatment failure of 351 patients enrolled on 3 consecutive protocols for childhood AML between 1991 and 2008 were analyzed according to age and protocol.
The more recent protocol (AML02) produced improved outcomes for patients aged 10 years to 21 years compared with 2 earlier studies (AML91 and AML97), with 3-year rates of event-free survival (EFS), overall survival (OS), and cumulative incidence of refractory leukemia or recurrence (CIR) for this group being similar to those of patients aged birth to 9 years: EFS: 58.3% ± 5.4% versus 66.6% ± 4.9% (P = .20); OS: 68.9% ± 5.1% versus 75.1% ± 4.5% (P = .36); and CIR: 21.9% ± 4.4% versus 25.3% ± 4.2% (P = .59). The EFS and OS estimates for patients aged 10 to 15 years overlapped those for patients aged 16 to 21 years. However, the cumulative incidence of toxic death was significantly higher for patients aged 10 to 21 years compared with younger patients (13.2% ± 3.6% vs 4.5% ± 2.0%; P = .028).
The recognition that cancer survival rates in adolescents have not improved as dramatically as those in younger children1, 2 has led to a growing interest in the field of adolescent and young adult oncology.3, 4 Data from the Surveillance, Epidemiology, and End Results (SEER) 9 study demonstrate that survival rates of younger children with acute myeloid leukemia (AML) increased considerably from the 1980s to the 1990s, whereas the corresponding rates for older children improved only modestly.1 Similarly, we previously found that, among children and adolescents with AML who were treated at St. Jude Children's Research Hospital and The University of Texas MD Anderson Cancer Center, survival rates improved between the 1980s and the 1990s for patients aged ≤ 10 years but not for those aged 10 years to 21 years.5 Among those treated in the 1990s, survival was significantly worse for older patients.5
Our recently reported study for children and adolescents with newly diagnosed AML (AML02; 2002-2008) yielded 3-year event-free survival (EFS) and overall survival (OS) estimates of 63% and 71%, respectively,6 indicating substantial gains in outcome compared with our previous trial (AML97; 1997-2002), with corresponding estimates of 44% and 50%.7 We attributed this improvement to the use of risk-adapted therapy based on sequential measurements of minimal residual disease as well as to vigilant supportive care, including the use of prophylactic antimicrobials. In the context of these changes in treatment strategy, we reassessed the effect of age on outcome, with an additional focus on the adolescent age group.
MATERIALS AND METHODS
Patients with AML who were aged ≤ 21 years and enrolled on 1 of 3 consecutive St. Jude AML protocols (AML91,8 AML97,7 and AML026) were included in these analyses. Briefly, AML918 included 1 or 2 courses of cladribine, 2 courses of DAV (daunorubicin at a dose of 30 mg/m2/day by continuous infusion on days 1-3, cytarabine at a dose of 250 mg/m2/day by continuous infusion on days 1-5, and etoposide at a dose of 200 mg/m2/day by continuous infusion on days 4 and 5), and allogeneic or autologous hematopoietic stem cell transplantation (HSCT). The AML977 trial was comprised of 1 course of cladribine plus cytarabine; 2 courses of DAV; and subsequent allogeneic HSCT, autologous HSCT, or chemotherapy. AML026 was a randomized trial of high-dose cytarabine (3 g/m2 every 12 hours on days 1, 3, and 5) or low-dose cytarabine (100 mg/m2 every 12 hours on days 1-10) plus daunorubicin (50 mg/m2 on days 2, 4, and 6) and etoposide (100 mg/m2 on days 2-6) given as induction I. Induction II was comprised of low-dose cytarabine, daunorubicin, and etoposide alone or in combination with gemtuzumab ozogamicin. Postremission therapy included allogeneic HSCT or chemotherapy, based on the risk of disease recurrence.
Patients with Down syndrome, acute promyelocytic leukemia, secondary AML, or biphenotypic leukemia were excluded. The protocols were approved by the Institutional Review Boards, and written informed consent and assent, as appropriate, were obtained for all patients.
OS was defined as the time elapsed from study enrollment until death, with those patients still alive at the date of last follow-up being censored. EFS was defined as the time elapsed from study enrollment to induction failure (refractory leukemia after 2 courses of therapy), withdrawal from the study, disease recurrence, secondary malignancy, or death, with those patients still alive and event free being censored at the time of last follow-up. In analyses of the cumulative incidence of induction failure and disease recurrence, withdrawal from the study, death, and secondary malignancy were considered to be competing events. In analyses of the cumulative incidence of toxic death, induction failure, disease recurrence, and secondary malignancy were considered to be competing events. For patients treated on AML02, the duration of each course of therapy was defined as the time elapsed from the start date of that course of therapy until the start date of the subsequent course of therapy.
The Kaplan-Meier method9 was used to estimate the probability of OS and EFS, and standard errors were determined by the method of Peto and Pike.10 Survival comparisons were made by performing the Mantel-Haenszel log-rank test with permutation P values determined by 10,000 Monte Carlo simulations. The Gray method was used to estimate and compare the cumulative incidences of induction failure, disease recurrence, and toxic death.11 The association between age group and duration of treatment was assessed using the exact Kruskal-Wallis test with Monte Carlo simulations. The association between age and duration of treatment was assessed using the Spearman correlation. All analyses were performed using SAS statistical software (Windows version 9.2; SAS Institute Inc, Cary, NC) and R software (Windows version 2.9.0; R Foundation for Statistical Computing, Vienna, Austria).
Among the 351 patients with de novo AML who were enrolled on the AML91, AML97, and AML02 protocols, 150 were aged 10 to 21 years (Table 1). The distributions by French-American-British (FAB) subtype (P < .001), cytogenetic group (P < .0001), and study protocol (P = .02) differed significantly among age groups. FAB M5 and M7 subtypes and 11q23 abnormalities were less common, whereas the M2 subtype and t(8;21) were found to be more common among patients aged 10 years to 21 years.
Abbreviations: AML91 indicates St. Jude acute myeloid leukemia (AML)91 protocol; AML97, St. Jude 97 AML protocol; AML02, St. Jude 02 AML protocol; FAB, French-American-British classification; Misc, miscellaneous; WBC, white blood cell count.
Determined using the Fisher chi-square test.
Effects of Treatment Protocol on Outcome
The 3-year EFS estimates were significantly higher for patients aged 10 years to 21 years who received AML02 therapy (92 patients) than for those treated on AML91 or AML97 (58 patients) (58.3% ± 5.4% vs 41.4% ± 6.3%; P = .041) (Fig. 1a). OS estimates were also significantly higher (68.9% ± 5.1% vs 48.3% ± 6.4%; P = .005) (Fig. 1b). These improvements were paralleled by correspondingly lower rates of induction failure and disease recurrence (3-year estimates: 21.9% ± 4.4% for AML02 vs 37.9% ± 6.5% for AML91 and AML97; P = .039).
The improvement in outcome obtained in AML02 extended to patients aged < 10 years; the 3-year EFS and OS estimates were significantly higher in AML02 (111 patients) than in AML91 or AML97 (90 patients): EFS: 66.6% ± 4.9% versus 50.0% ± 5.2% (P = .013) (Fig. 1c) and OS: 75.1% ± 4.5% versus 60.0% ± 5.1% (P = .019) (Fig. 1d). The 3-year cumulative incidence of induction failure or disease recurrence estimates were found to be lower (25.3% ± 4.2% vs 40.0% ± 5.2%; P = .025).
In contrast to the improvements in the EFS, OS, and disease recurrence rates observed in AML02, there was only a modest decrease in the cumulative incidence of toxic death noted among patients treated on AML02 (8.4% ± 2.0%) compared with that of patients treated on the earlier protocols (12.2% ± 2.7%) (P = .18). For patients aged 10 years to 21 years, the 3-year cumulative incidence of toxic death was 13.2% ± 3.6% on AML02, compared with 19.0% ± 5.2% on the earlier studies (P = .23). For patients aged < 10 years, the rates of toxic death were 4.5% ± 2.0% on AML02 and 7.8% ± 2.8% on AML91 and AML97 (P = .33).
Comparisons Between Age Groups in AML02
Among the patients enrolled on AML02, there was no significant difference noted in EFS between those who were aged 10 years to 21 years at the time of diagnosis (92 patients) and younger patients (111 patients) (3-year estimates: 58.3% ± 5.4% vs 66.6% ± 4.9% P = .20) (Fig. 2a) (Table 2). Similarly, there was no significant difference noted with regard to OS or the cumulative incidence of induction failure or disease recurrence between the age groups (3-year estimates: 68.9% ± 5.1% vs 75.1% ± 4.5% [P = .36] and 21.9% ± 4.4% vs 25.3% ± 4.2% [P = .59], respectively) (Figs. 2b and 2c) (Table 2). In a competing risk regression model that treated age as continuous variable, older age was not found to be associated with induction failure or disease recurrence (hazard ratio [HR], 0.984; 95% confidence interval [95% CI], 0.944-1.027 [P = .46]).
Table 2. Outcome and Causes of Failure Among Patients Treated on AML02
In contrast to the similarities in EFS, OS, and the cumulative incidence of induction failure or disease recurrence between the 2 age groups, the 3-year cumulative incidence of toxic death of 13.2% ± 3.6% for patients aged 10 years to 21 years was significantly higher than the incidence of 4.5% ± 2.0% noted for younger patients (P = .028) (Fig. 2d) (Table 2). This difference in toxic death rates according to age was also observed among patients treated on the AML91 or AML97 protocols (19.0% ± 5.2% for patients aged 10 years-21 years compared with 7.8% ± 2.8% for younger patients; P = .023).
Within AML02, a competing risk regression model that included Fms-like tyrosine kinase 3 (FLT3) status (internal tandem duplication vs other) and karyotype (favorable vs other) revealed that the cumulative incidence of toxic death of patients aged 10 years to 21 years differed significantly from that of patients aged birth to 9 years (HR, 3.303; 95% CI, 1.06-10.30 [P = .039]). When age was treated as a continuous variable, older age was found to be marginally associated with an increased risk of toxic death after adjusting for FLT3 status and karyotype (HR, 1.111; 95% CI, 0.997-1.237 [P = .057]).
Among the 17 patients who died of causes other than refractory leukemia or disease recurrence, 12 died of infection because of bacterial (6 patients), fungal (3 patients), or viral (3 patients) pathogens. Twelve patients were aged 10 years to 21 years; the causes of death in this age group included infection (9 patients), hemorrhage (2 patients), and encephalopathy of unknown cause (1 patient). Of the remaining 5 patients who died of toxicity, 4 were infants who died of infection (2 patients), hemorrhage (1 patient), and narcotic overdose (1 patient). To investigate potential causes of the increased risk of infection among older patients, we analyzed the length of each course of therapy according to age group. The duration of consolidation therapy I was significantly longer for patients aged 10 years to 21 years, with a mean of 35 days (range, 22 days-149 days) versus 32 days (range, 21 days-86 days) for younger patients (P = .011). The difference in duration was even greater for consolidation therapy II: 49 days (range, 35 days-135 days) versus 41 days (range, 22 days-69 days) for younger patients (P < .001).
Finally, we performed additional analyses of patients who were aged 16 years to 21 years to determine whether their outcome differed from that of children aged 10 years to 15 years. The 3-year EFS and OS estimates were nearly identical between patients aged 10 years to 15 years (EFS: 59.2% ± 6.1% and OS: 69.6% ± 4.7%) and patients aged 16 years to 21 years (EFS: 56.7% ± 9.6% and OS: 67.2% ± 9.3%). In addition, the cumulative incidence of induction failure or disease recurrence (23.5% ± 5.4% and 17.9% ± 7.4%) and the cumulative incidence of toxic death (12.5% ± 4.2% and 14.8% ± 7.1%) were similar between the patients aged 10 years to 15 years and those aged 16 years to 21 years.
In the current study, we investigated the impact of recent advances in treatment on the outcome of patients aged 10 years to 21 years with pediatric AML. The results of this study demonstrate that OS rates increased by approximately 20% and disease recurrence rates decreased by approximately 15% for patients in this age group who were treated between 2002 and 2008 compared with those who were treated from 1991 through 2001. In addition, rates of OS, EFS, and the cumulative incidence of refractory disease or disease recurrence for older patients treated on AML02 were not significantly inferior to those of younger patients. Among patients treated on AML02, the only age-associated difference in outcome was the increased incidence of toxic deaths among patients aged 10 years to 21 years. Treatment-related mortality was higher among patients aged 10 years to 21 years than among patients aged birth to 9 years on both the older (AML91 and AML97) and more recent (AML02) protocols. While AML02 was being conducted, we implemented a series of supportive care measures among patients treated at our institution.12 Although we demonstrated that the use of prophylactic vancomycin, ciprofloxacin, and voriconazole significantly reduced the incidence of bacteremia and the length of hospitalization, this measure did not appear to impact treatment-related mortality.12 It should be noted, however, that these supportive care guidelines were introduced approximately 2 years after accrual to the study began and were not implemented at all sites. It is conceivable that strict adherence to these guidelines may affect toxic deaths in future trials. Nevertheless, in the AML02 protocol, toxic deaths, primarily from infection, still occurred in > 10% of these patients.
Results of our previous study demonstrated that age was an independent prognostic factor in childhood AML, with lower EFS and OS rates and a higher cumulative incidence of refractory disease or disease recurrence noted in patients aged at least 10 years.5 In that study, which included patients who were treated from 1983 through 2002, the effect of age was noted primarily among patients treated between 1990 and 2002, during which time the outcome improved for younger children but not for older children. Similarly, in an analysis of patients with AML who were treated on 6 trials from 1993 through 2004, Creutzig et al13 reported that older age was an independent negative predictor of outcome. These investigators found that patients aged 2 years to 12 years had higher rates of complete remission and lower disease recurrence rates than did patients in other age groups. In addition, there was a trend toward higher treatment-related mortality with increasing age.13 In a study of AML patients treated from 1986 to 1999, Horibe et al14 demonstrated that children aged 1 year to 9 years had the highest EFS rate (43%), with lower EFS rates reported for older age groups (34% for patients aged 10 years-15 years, 32% for patients aged 15 years-19 years, and 26% for those aged 20 years-29 years).
A variety of reasons have been set forth to explain the inferior survival rates for adolescents and young adults with cancer, including delays in diagnosis, insurance barriers, low rates of enrollment on clinical trials, poor adherence, high-risk biological features, and increased treatment-related toxicity.2, 3 The majority of these factors, however, do not apply to the older patients analyzed in the current study, all of whom were enrolled in clinical trials for AML, and received the planned protocol-directed therapy regardless of insurance status. Poor adherence is unlikely to be a major issue in patients with AML because all chemotherapy is given intravenously, with most patients receiving their therapy in the inpatient setting. The results of the current study indicate that, in the context of intensive risk-based therapy with sequential minimal residual disease monitoring, adolescents with AML no longer experience inferior survival rates.
To the best of our knowledge, there is currently no clear explanation for the increased toxicity of AML treatment noted in older pediatric patients, which has been previously reported.15-18 Adolescents with AML typically do not have the comorbidities associated with toxicity in older adults and are rarely receiving concomitant medications that may increase toxicity. It is possible that subtle differences in immunologic function, coagulation, or pharmacokinetics may contribute to the differences in toxicity. In addition, the increased length of each course of consolidation therapy in older patients makes prolonged neutropenia a potential contributing factor to the increase in infection-related deaths. Conceivably, cumulative hematologic toxicity could be greater in older patients because of multiple factors, including slower clearance of chemotherapeutic agents, decreased bone marrow reserve, or reduced telomere length in hematopoietic precursors. Although the routine use of granulocyte–colony-stimulating factor does not decrease the risk of infectious complications or affect outcome in children with AML,19 its effect in the adolescent age group remains to be determined.
The results of the current study have demonstrated that the outcome of adolescents with AML has improved in the recent treatment period and that the rates of recurrent and refractory disease in adolescents are similar to those in younger patients. Current trials must focus on elucidating the causes of increased toxicity and improving supportive care for older children with AML.
We thank Cherise Guess for expert editorial review, Kathy Jackson and Heidi Clough for data collection, and Julie Groff for preparing the figures.
Note Added in Proof
Supported in part by Cancer Center Support (CORE) grant P30 CA021765-30 from the National Institutes of Health and by the American Lebanese Syrian Associated Charities (ALSAC).
CONFLICT OF INTEREST DISCLOSURES
Ching-Hon Pui is an American Cancer Society Professor.