A retrospective meta-analysis of adolescents and young adults (AYAs) with acute myeloid leukemia (AML) was performed to determine if differences in outcome exist following treatment on pediatric versus adult oncology treatment regimens.
A retrospective meta-analysis of adolescents and young adults (AYAs) with acute myeloid leukemia (AML) was performed to determine if differences in outcome exist following treatment on pediatric versus adult oncology treatment regimens.
Outcomes were compared of 517 AYAs with AML aged 16 to 21 years who were treated on Children's Oncology Group (COG), Cancer and Leukemia Group B (CALGB), and Southwest Oncology Group (SWOG) frontline AML trials from 1986 to 2008.
There was a significant age difference between AYA cohorts in the COG, CALGB, and SWOG trials (median, 17.2 versus 20.1 versus 19.8 years, P < .001). The 10-year event-free survival of the COG cohort was superior to the combined adult cohorts (38% ± 6% versus 23% ± 6%, log-rank P = .006) as was overall survival (45% ± 6% versus 34% ± 7%), with a 10-year estimate comparison of P = .026. However, the younger age of the COG cohort is confounding, with all patients aged 16 to 18 years doing better than those aged 19 to 21 years. Although the 10-year relapse rate was lower for the COG patients (29% ± 6% versus 57% ± 8%, Gray's P < .001), this was offset by a higher postremission treatment-related mortality of 26% ± 6% versus 12% ± 6% (Gray's P < .001). Significant improvements in 10-year event-free survival and overall survival were observed for the entire cohort in later studies.
Patients treated on pediatric trials had better outcomes than those treated on adult trials, but age is a major confounding variable, making it difficult to compare outcomes by cooperative group. Cancer 2013;119:4170–4179. © 2013 American Cancer Society.
Acute myeloid leukemia (AML) affects patients of all ages, and survival rates in general decrease with advancing age; many factors might contribute to this fact. Adolescents and young adults (AYAs) with AML are cared for by both pediatric and adult oncologists, with dose intensity higher in pediatrics. AYAs with acute lymphoblastic leukemia (ALL) have improved survival when treated on pediatric treatment regimens compared to regimens designed for older adults.[1, 2]
Two small preliminary studies have addressed whether AYAs have better outcomes on pediatric or adult AML trials. AYAs treated on one Children's Cancer Group (CCG) trial did better than AYAs treated with adult therapies at the MD Anderson Cancer Center (MDACC), but the CCG patients were younger than those at MDACC. Patients younger than 21 years treated at St. Jude Children's Research Hospital and MDACC demonstrated that survival decreased with advancing age, with minimum examination of results by treatment regimens.
This report compares the relative effectiveness of adult and pediatric AML therapy in AYA patients using data from Children's Oncology Group (COG), Cancer and Leukemia Group B (CALGB), and Southwest Oncology Group (SWOG) frontline studies.
All patients provided informed consent according to federal and institutional guidelines and in accordance with the Declaration of Helsinki. All COG, CALGB, and SWOG studies included in this analysis were performed using protocols approved by institutional review board. The later studies were registered at www.ClinicalTrials.gov since its inception in 2000.
A total of 281 patients aged 16 to 21 years were enrolled on CCG-2861, CCG-2891,[3, 7] CCG-2941, CCG-2961, and COG AAML03P1 studies from 1986 to 2008, with details in the references and Table 1. The first 4 trials each used an “intensive-timing” induction, and for postremission therapy compared outcomes of patients receiving aggressive high-dose cytarabine, and in 2 cases autologous transplantation, to those assigned to allogeneic blood or marrow transplantation (BMT). In COG AAML03P1, induction was intensified using a UK Medical Research Council approach.
|Protocol/Courses (Reference)||Induction Courses, Drugs and Timing||Postremission Courses, Drugs, and Timing||BMT|
|COG-2861 + COG-2891 (3,8–9)Intensive timing, 6 courses||Dexamethasone, AraC, 6 thioguanine, VP, and Rubidomycin (DNM) (DCTER), given over 4 days (cycle 1) followed by a 6-day rest, then repeated over another 4 days (cycle 2), ×2 (ie, 2nd induction course repeated in an identical fashion after CBC recovery)||1st Course: timing intensive HiDAC, "Capizzi II", bid total 8 doses days 1,2 and 8,9, with L-asparaginase2nd and 3rd Courses: 28 day cycles of 6 thioguanine, vincristine, AraC, cyclophosphamide, and 5 Azacytidine4th Course: DCTER 1 cycle (4 days) only||Allogeneic if MFD; Others allocated to autologous BMT [COG-2861] or randomized to chemo vs autologous BMT [COG-2891]|
|COG-2941 + COG-2961 (10,11) 3 Courses||Intensively timed DCTER as above except first cycle in each of the 2 courses substituted idarubicin for DNM||Capizzi II as per 2861/2891[patients were then randomized to no further therapy or intravenous IL-2, with no differences in outcome].||Allogeneic if MFD available|
|COG-03P1 (12) 5 Courses||AraC, DNM, and VP as a 10-day course [ADE10] plus GO2nd course ADE over only 8 days||1st Course: HiDAC bid ×5 days, and etoposide ×5 days2nd Course: HiDAC bid days 1–4, mitoxantrone days 4–7 and GO3rd Course: Capizzi II as per 2861/2891||Allogeneic if MFD available|
|CALGB 8525 (14) 5 to 6 Courses||AraC and DNM ("7+3")x1If patients had morphologic residual disease after Course #1, a second Course of "5+2" was given on day 14||4 Courses of 3 randomized arms of AraC: 100 mg/m2/d ×5, 400 mg/m2/d ×5, or 3 g/m2 bid ×6 given on day 1, 3, and 5||None|
|CALGB 9022 (15) 5 to 6 Courses||Same as CALGB 8525||3 Courses: 1) HiDAC 3 g/m2 bid given on days 1, 3, and 52) VP/cyclophosphamide ×3 days3) AZQ and mitoxantrone ×3 days||None|
|CALGB 9222 (16) 5 to 6 Courses||Same as CALGB 8525||2 Randomized arms: 1) HiDAC 3 g/m2 bid given on days 1, 3, and 5: repeated ×32) HiDAC as above, followed by VP/cyclophosphamide as in CALGB 9022, followed by AZQ and mitoxantrone as in CALGB 9022||None|
|CALGB 9621 (17) 2 to 5 Courses||AraC (7 days), DNM (3 days), VP (3 days), or ADE randomized to receive or not receive PSC-8333 (Valspodar) as an MDR1 inhibitor If residual disease, a second course of AraC (5), DNM (2), and VP was given on day 14||2 Arms1) Core binding factor (CBF) AML: HiDAC as in 9022 repeated ×32) others: autologous BMT||Autologous except CBF + AML|
|CALGB 19808 (18) 3 to 5 Courses||Same as 9621, randomized to receive or not receive PSC-833||2 Arms1) CBF AML HiDAC as in 9022 x32) 2-step autologous BMT: HiDAC and HD VP for in vivo purging, followed by Autologous BMT||Same as CALGB 9621|
|SWOG S58600 (19) 2 to 4 Courses||Randomized Arms1) "7+3" ×12) HiDAC bid ×6 days, DNM days 7–9, ×1If residual AML: Repeat same regimen ×1||Induction 1) Randomized armsA) HiDAC bid ×5 days,DNM days 6–7 ×1B) "7+3", 2 coursesInduction 2) Postremission armA) ×1||None|
|SWOG S9500 (20) 2 to 9 Courses||"7+3" followed by HiDAC bid days 8–10, ×1||HiDAC bid days 1, 3, and 5 as tolerated to 4 courses, followed by "5+1" as tolerated to 4 courses||None|
|SWOG S0106 (21) 4 to 7 Courses||Randomized arms"7+3" plus GO day 4, ×1>2) "7+3" ×1If residual AML on day 14: "7+3"||HiDAC bid days 1, 3, and 5, ×3 followed by randomization: GO ×3 vs none||Allogeneic, only for patients with adverse cytogentics and matched sibling donor|
During 1986 to 2008, 149 patients aged 16 to 21 years were enrolled on sequential CALGB trials for newly diagnosed AML, including CALGB 8525, CALGB 9022, CALGB 9222, CALGB 9621, and CALGB 19808. All used daunorubicin/cytarabine-based induction and high-dose cytarabine–based intensification. In the CALGB 9621 and CALGB 19808 trials, autologous transplantation was performed in patients without core-binding factor cytogenetics, with no allogeneic transplants offered. SWOG enrolled 87 AYAs on 3 frontline trials for AML from 1986 to 2008: SWOG 8600, SWOG 9500, and SWOG S0106. All studies used daunorubicin and cytarabine, and in one case, this therapy plus gemtuzumab ozogamicin. Neither autologous nor allogeneic transplantation were offered except that allogeneic transplantation be considered for patients with high-risk cytogenetics and a matched sibling donor in SWOG S0106.
CCG-2861 was current as of September 21, 2001, CCG-2891 was current as of January 14, 2004; CCG-2961, November 6, 2009; CCG-2941, April 14, 2005; and COG AAML03P1, May 12, 2010. Data from CALGB and SWOG studies were current as of June 28, 2010, and May 12, 2010, respectively. The significance of observed differences in proportions was tested using the chi-square test and Fisher's exact test when data were sparse. The Mann-Whitney test was used to determine the significance between differences in median values. Study entry characteristics analyzed included sex, white blood cell count, bone marrow blasts percentage, FAB (French-American-British) classification, and weight-related groups defined by body mass index (BMI) percentage. Median times from diagnosis to study entry for the COG, CALGB, and SWOG studies were 1, 1, and 2 days, respectively. Race and ethnicity were not analyzed due to differences in data collection among the groups. Weight groups were defined as either underweight (BMI < 11%), middleweight (11%-94%), or overweight (BMI ≥ 95%) using accepted American standards. For cytogenetics, patients were classified as per the Byrd classification system as favorable [t(8;21), inv(16) or t(16;16)], adverse [complex karyotype ≥ 3 abnormalities, inv(3) or t(3;3), t(6;9), t(6;11), −7, +8 sole or with one other abnormality not favorable, or t(11;19)], or intermediate (all others).
Overall survival (OS) was defined as time from study entry to death. Event-free survival (EFS) was defined as time from study entry until death, relapse, or failure to achieve complete remission (CR) after receiving up to 2 courses of induction therapy, except for patients who enrolled on SWOG S9500 who received only one course of induction. Relapse-free survival (RFS) was defined as time from end of induction (EOI) for patients in CR, censoring patients who died without an intervening relapse. Relapse risk (RR) was measured from end of induction for patients in CR to relapse where deaths without a relapse were considered competing events. Postremission treatment-related mortality (TRM) was recorded from EOI for patients in CR to death without a relapse, censoring relapses. AYAs lost to follow-up were censored at their date of last known contact. Patients who received allogeneic BMTs were not censored in these analyses, with the exception of the indicated RR and TRM, and some OS and EFS analyses in which COG patients were censored at the time of study transplant because few patients on the adult trials received an allogeneic transplant and were unidentified.
The Kaplan-Meier method was used to estimate OS and EFS. Estimates include 95% confidence intervals (CIs) using Greenwood's estimate of the standard error. Cumulative incidence of TRM and RR were estimated using the method of Kalbfleisch and Prentice. The significance of predictor variables was tested with the log-rank statistic for OS and EFS. Gray's statistic was used to compare cumulative incidence curves for RR and TRM. Cox proportional hazard models were used to estimate hazard ratios (HR) for cohorts of patients defined by age and cooperative group and for analyzing age as a continuous variable for univariate and multivariate analyses of OS, EFS, and RFS. The proportional hazards assumption was tested for all covariates. Analyses of OS that compared 16- to 18-year-old patients versus 19- to 21-year-old patients or patients from COG versus CALGB and SWOG studies violated the proportional hazards assumption, and therefore a direct comparison between the 10-year estimates of OS were summarized instead of the log-rank statistic.
Patient characteristics are shown in Table 2. Overall, there were 517 patients included in the analysis, with a median age of 18 years, with the range in all 3 groups from 16 to 21 years. However, the COG cohort was significantly younger with a median age of 17.2 years, P < .001, with 94% of the COG patients younger than 19 and only 1 patient at 21 years. The median ages of the CALGB (20.1 years) and SWOG (19.8 years) patients were comparable; only 4 were 16 years old.
|All Patients (N = 517)||COG Studies (N = 281)||CALGB Studies (N = 149)||SWOG Studies (N = 87)|
|Age, y, median (range)||18.0||(16.0–21.99)||17.2||(16.0–21.6)||20.1||(16.5–21.98)||19.8||(16.9–21.99)||<.001|
|White blood cell (×103/μL), median (range)||19.1||(0.5–860)||20||(0.5–860)||18.1||(0.5–453)||16.8||(0.6–415.8)||.669|
|Bone marrow blasts (%), median (range)||73||(0–100)||74||(1–100)||71.5||(0–99)||70.5||(0–100)||.570|
|ALL Patients (N = 517)||COG Studies (N = 281)||CALGB Studies (N = 149)||SWOG Studies (N = 87)||P|
|Weight group (body mass index [BMI])|
|Underweight (BMI ≤11%)||52||10%||24||9%||20||13%||8||9%||.103|
|Middleweight (BMI 11%-94%)||346||67%||201||72%||90||60%||55||63%||.806|
|Obese (BMI ≥ 95%)||117||23%||54||19%||39||26%||24||28%||.233|
The median white blood cell count and blast percentage at diagnosis were 19.1 × 109/L and 73%, respectively, with no differences by cooperative group in these or in the FAB classification distribution. Cytogenetic data are reported in Table 3. There was an even distribution across the 3 cooperative groups in the proportion of patients in each risk group for all endpoints except for a paucity of patients in the adverse risk group from SWOG. Cytogenetic results were unknown on almost half the patients. Molecular data were incomplete during the study period, especially in the early years, and were not analyzed.
|Risk Group||All Patients N = 517||COG Studies N = 281||CALGB Studies N = 149||SWOG Studies N = 87||P|
|Risk Group for Complete Response|
|Risk Group for Relapse|
|Risk Group for Overall Survival|
The CR rate after receiving up to 2 courses of induction was 79% for the entire cohort. These rates were significantly different among all 3 groups (COG, 82%; CALGB, 76%; SWOG, 71%; P = .045), and there were no differences in actuarial survival at 60 days (92%, 95%, and 95%, respectively).
Ten-year OS was higher for the COG cohort (Fig. 1A) than for the 2 adult cohorts, 45% ± 6% versus 34% ± 7% with a 10-year estimate comparison of P = .026. The adult trials had similar OS, 35% ± 8% for CALGB and 33% ± 12% for SWOG, P = 1.00. Similarly, the 10-year EFS was 38% ± 6% for the 281 AYAs on COG trials compared to 23% ± 6% for the patients on the adult trials (log-rank P = .006; Fig. 1B). Results from CALGB and SWOG were comparable, at 24% ± 8% and 21% ± 10%, respectively; hence, for all other analyses, the adult groups were combined for a single comparison to COG. When we repeated the OS and EFS analyses censoring the COG transplant recipients (N = 77), the overall results were similar. EFS for COG chemotherapy-only patients was 36% ± 7%, P = .023 versus 23% on adult trials; and OS 44% ± 7%, P = .053 versus 34% on adult trials.
In examining age, OS at 10 years for the 16- to 18-year-old patients was 43% ± 6% compared with 32% ± 8% for those aged 19 to 21 years (P = .034) when comparing 10-year estimates (Fig. 1C). Figure 1D shows a significant difference in EFS between the 2 age cohorts regardless of treatment, with patients 16 to 18 years old (N = 341) having a 10-year EFS of 34% ± 6% compared to 21% ± 7% for those aged 19 to 21 years (N = 176; log-rank P = .015). When one stratifies the entire cohort based on pediatric versus adult protocols and age (Fig. 2A) for OS, there were no significant differences among the 4 curves.
AYAs treated on the COG protocols had a markedly reduced incidence of relapse at 10 years, with 29% ± 6%, and 35% ± 8% censoring BMT patients, compared to those treated on adult trials, 57% ± 8% (Gray's P < .001 for both comparisons). Younger patients on both the pediatric and adult trials had fewer relapses, at 34% ± 6% versus 58% ± 10% for the older AYAs (Gray's P < .001). Both age groups treated on COG trials had much lower relapse rates than patients on the adult trials regardless of age (Fig. 2B).
Because of differences in the definition of TRM among the cooperative groups during induction therapy, we analyzed TRM only from EOI. AYAs treated on the more myelosuppressive COG protocols had a higher degree of TRM (26% ± 6%) than those treated on the adult protocols (12% ± 6%, Gray's P < .001). Censoring BMT patients on COG trials only reduced the TRM to 22% ± 7%, still P < .001 compared to adult trials. Figure 2C shows the TRM stratified by both cooperative groups and age (Gray's P < .001 overall). Age played less of a role in TRM compared to the dramatic differences seen in RR.
Patient characteristics were examined as potential confounders to the superior outcomes of patients on pediatric trials. Cox linear regression analyses were performed using age as a continuous variable examining endpoints (Table 4). No significant differences were seen for OS. However, increasing age was a significant risk factor for EFS and RFS, despite a lower TRM associated with increasing age. Looking at the adult and pediatric groups individually, only EFS in the COG studies demonstrated poorer outcomes with increasing age (HR = 1.16, P = .038). No other patient characteristics examined were prognostic except for adverse cytogenetics, which was highly significant (P < .001). There was a lower incidence of TRM for patients with favorable cytogenetics on both pediatric and adult trials (HR = 0.53, P = .014).
|Age in years: continuous variable||HR||95% CI||P|
|Overall survival from study entry|
|Event-free survival from study entry|
|Treatment-related mortality from remission (N = 396)|
|Relapse-free survival from remission (N = 396)|
Univariate and multivariate analyses were done for EFS from study entry looking at age in 2 discrete cohorts as above, and pediatric versus adult studies (Table 5). Increased age and being on adult studies were risk factors for outcome. However, in the multivariate analyses, neither age nor studies used showed a significant difference, because there was such a strong correlation between age and pediatric versus adult trials. Multivariate analysis of OS was not appropriate due to nonproportional hazards. Finally, we looked at just the 16- to 18-year-old cohort comparing patients treated on either COG or CALGB/SWOG trials. OS for those treated on the COG protocols (N = 263) was 44.9% ± 6.6% at 10 years versus 39.5% ± 11.6% for the 78 patients treated on the adult trials (P = .417). There remained a significant reduction in relapse risk for patients treated on the COG protocols, whereas the TRM was significantly higher for the same patients on COG trials.
|Event-Free Survival From Study Entry||Cox Analyses|
|Age: 16–18 y||341||1.00|
|Age: 19–21 y||176||1.31||1.05–1.63||.015|
|Age in years (continuous)||1.06||0.97–1.15||.238|
|Age: 16–18 y||1.00|
|Age: 19–21 y||1.13||0.85–1.50||.408|
Multivariate analyses were run to determine if the period of study (early, 1986-1995; late, 1996-2008) or cytogenetics played any confounding role. Only adverse cytogenetics was confounding in the limited subset of patients having known cytogenetics. After adjusting for cytogenetics in multivariate models, 19- to 21-year-olds (HR = 1.24, 95% CI = 0.92-1.68, P = .165) and patients on adult studies (HR = 1.32, 95% CI = 0.99-1.77, P = .062) had nonsignificantly worse EFS.
Finally, we compared the outcomes of studies before 1996 to studies from 1996 forward. OS at 10 years increased from 34% ± 6% to 48% ± 6% (P = .045); and EFS from 26% ± 6% to 33% ± 8% (P = .039) in later studies. Unfortunately, TRM doubled from 12% ± 5% to 26% ± 6% (Gray's P < .001), but the relapse rate markedly declined, from 51% ± 8% to 36% ± 10% in more recent studies (Gray's P < .001).
There has been increasing attention given to the outcome of AYAs with cancer. In the United States, lesser improvements in survival have been found in younger adults compared to either children or older adults. This fact may be driven in part by a much lower participation rate in clinical trials, as noted for some cancers. However, treatment effects probably play the largest role. As noted, AYA patients with ALL will do substantially better if treated on pediatric protocols.[1, 2] Patients in this age group with common pediatric tumors such as rhabdomyosarcoma and Ewing's sarcoma also seem to fare better when treated on pediatric protocols. However, for cancers common in both age groups, eg, Hodgkin lymphoma, results are comparable. AYAs with adult-type cancers may actually fare better in the hands of adult oncologists.
Small preliminary studies in de novo AML were inconclusive regarding the optimum approach of treating AYAs on childhood or adult protocols.[4, 5] Age was found to be an important prognostic factor in patients aged 0 to 55 years treated on a common UK Medical Research Council protocol. When pediatric-like therapy was administered to adult patients younger than 50 years, increased death from toxicity counterbalanced an improvement in leukemia-free survival.
We investigated AML outcomes of 517 AYAs from 3 large cooperative groups (COG, CALGB, and SWOG) making this the largest attempt at comparing pediatric and adult therapy. We took into account all potential measurable variables that could skew results. With the exception of age, none of the other characteristics seemed to play a role in the results obtained. In the limited subset of patients with adequate cytogenetics, multivariate analyses revealed that adverse karyotypes were an important prognostic factor, but hazard ratios for older age (1.24) and adult studies (1.32) remained high, albeit with limited power. The OS and EFS were superior for patients treated on COG protocols, but significantly more patients on the adult trials were older, and those patients did worse than younger adolescents regardless of protocol. Even in this small age range of 6 years, the influence of increasing age on lowering survival rates was noteworthy. We found this fact most surprising, but appeared to be from both higher relapse rates and higher TRM among the 19 to 21 year olds. Our best explanation is that this 6-year period is a microcosm of overall results in AML between children and older adults.
However, we noted striking differences between causes for mortality. The more myelosuppressive pediatric protocols had more anti-leukemia efficacy than the adult trials, with halving of relapse rates; but the marked toxicity that ensued was also clearly apparent with TRM of 26% compared to 12% for the adult trials (P < .001).
Aggressive pediatric AML protocols are better tolerated in young children, with better survival than adults enrolled on trials designed for middle-aged adults, which must modify therapy for tolerability. Pediatric trials would be superior for AYAs if the TRM with current therapy could be lowered. Molecular markers and inhibitors are allowing treatment to be further stratified, sometimes with greatly improved outcome without major myelosuppression. But for the vast majority of patients with AML, it is not yet feasible to reduce profoundly myelosuppressive therapy and obtain optimal cure rates. In one of the pediatric trials cited herein, CCG-2961, a dramatic improvement in overall TRM resulted when an amendment requiring specific mandatory supportive care measures was implemented. Furthermore, the improvement in TRM specifically among the AYAs on CCG-2961 was even more dramatic, 43% for preamendment versus 22% postamendment (P = .03). Overall OS improved from 43% to 57%, and the results also documented a “learning curve,” reflected by a lowering of TRM with time, even before the supportive care amendment was implemented. This has been noted in adult trials, where traditionally supportive care is left to institutional guidelines, with results generalizable to community/standard practice. It is recommended that adult AML patients be cared for by physicians who are experienced in treating leukemia. The pediatric protocols have in general been more specific in outlining such guidelines.
Until more highly effective molecular inhibitors of specific AML subtypes are discovered, pediatric and adult oncologists taking care of these patients should focus their attention on supportive care measures to lower TRM. Perhaps an intergroup trial of AML in 16- to 30-year-olds could be implemented to better understand age-related differences in outcome, laying the groundwork for future AYA trials.
Generously funded by a grant from the Young Adult Alliance of the LIVESTRONG Foundation, The Coleman Leukemia Research Foundation; and grants U10_CA98543, CA98413, CA101140, CA77658, CA31946, CA41287, CA32102, and CA38926 from the National Cancer Institute (NCI).
Dr. Woods has received grants (cooperative group agreements) from the NCI. Dr. Franklin has received grants from LIVESTRONG. Dr. Donohue has received grant CA33601 from the NCI. All other authors made no disclosure.