Successes and challenges in the treatment of pediatric acute myeloid leukemia: a retrospective analysis of the AML-BFM trials from 1987 to 2012

Abstract

Overall survival (OS) of pediatric patients with acute myeloid leukemia (AML) increased in recent decades. However, it remained unknown whether advances in first-line treatment, supportive care, or second-line therapy mainly contributed to this improvement. Here, we retrospectively analyzed outcome and clinical data of 1940 pediatric AML patients (younger than 18 years of age), enrolled in the population-based AML-BFM trials between 1987 and 2012. While 5-year probability of OS (pOS) increased from 49 ± 3% (1987-1992) to 76 ± 4% (2010-2012; p < 0.0001), probability of event-free survival only improved from 41 ± 3% (1987-1992) to 50 ± 2% (1993-1998; p = 0.02) after introduction of high-dose cytarabine/mitoxantrone, but remained stable since then. Non-response and relapse rates stayed constant despite intensified first-line therapy (p = 0.08 and p = 0.17). Reduced fatal bleedings and leukostasis translated into fewer early deaths (8.1%vs. 2.2%; p = 0.001). Strikingly, pOS after non-response (13 ± 5% (1987-1992) vs. 43 ± 7% (2005-2010); p < 0.0001) or relapse (19 ± 4% vs. 45 ± 4%; p < 0.0001) improved. After 1999, more relapsed or refractory patients underwent hematopoietic stem cell transplantation (HSCT) with increased pOS after HSCT (29 ± 5% (1993-1998) vs. 50 ± 4% (2005-2010); p < 0.0001). Since efficacy of salvage therapy mainly contributed to better outcome in pediatric AML, our analysis indicates that a better allocation of patients, who cannot be cured with conventional chemotherapy, to an early "salvage-like" therapy is necessary.

Fig. 1

Development of overall and event-free survival from 1987–2012. a Protocol flow chart for pediatric patients with AML in AML-BFM trials from 87 until 2012. Consolidation: 6-week therapy consisting of seven different drugs: 6-thioguanine (60 mg/m² per day, days 1–43 orally); prednisone (40 mg/m² per day, days 1–28 orally); vincristine (1.5 mg/m² per day, days 1, 8, 15, 22); cytarabine (75 mg/m² per day, days 3–6, 10–13, 17–20, 24–27, 31–34, 38–41); doxorubicin (30 mg/m² per day (AML-BFM 87 and 93)) or idarubicin (7 mg/m² per day [AML-BFM 98], days 1, 8, 15, 22); intrathecal (age-dependent dose) cytarabine (days 1, 15, 29, 43); cyclophosphamide (500 mg/m² per day, days 29, 43). *CNS irradiation was stopped in May 2009. Cumulative dosages of anthracycline (mg/m²) and cytarabine (g/m²) in AML-BFM trials are shown as maximal dosage given in each study. Cumulative doses were calculated as equivalent doses to daunorubicin using a ratio of 1:5 for idarubicin and mitoxantrone. ADE cytarabine (100  mg/m2), daunorubicin (60  mg/m2), etoposide (150  mg/m2), AIE cytarabine (100  mg/m2), idarubicin (12  mg/m2), etoposide (150  mg/m2), HAM high-dose cytarabine (3  g/m2), mitoxantrone (10  mg/m2), AI cytarabine (500  mg/m2), idarubicin (7  mg/m2), hAM intermediate-dose cytarabine (1  g/m2), mitoxantrone (10  mg/m2), HAE high-dose cytarabine (3  g/m2), etoposide (125  mg/m2), ADxE cytarabine (100  mg/m2), liposomal daunorubicin (80 mg/m2), etoposide (150  mg/m2), 2-CDA 2-chloro-2-deoxyadenosine (6  mg/m2), CNS central nervous system. b Development of survival per 6-year periods. Data shown as probability of EFS and OS ± SE. Kaplan–Meier curves of EFS and OS of distinct subgroups were compared using the log-rank test, shown as p value. c Development of survival per 6-year periods in patients with standard risk group [FAB M1/2 with Auer rods, FAB M4 with atypical eosinophils (M4Eo) and/or favorable cytogenetics, such as t(8;21) and/or AML1-ETO and inv(16) or t(16;16) and/or CBFB/MYH1, if there was no persistence of BM blasts (≥5%) on day 15, respectively), compared to all other patients (high-risk group). Data shown as probability of EFS and OS ± SE. Kaplan–Meier curves of EFS and OS of distinct subgroups were compared using the log-rank test, shown as p value. Standard risk: n = 88 (1987–1992); n = 174 (1993–1998); n = 151 (1999–2004); n = 148 (2005–2010); n = 39 (2011–2012). High-risk: n = 207 (1987–1992); n = 359 (1993–1998); n = 346 (1999–2004); n = 329 (2005–2010); n = 98 (2011–2012). d Overview of events (early death, death in CR, non-response, relapse, others) per 6-year periods. Data shown as probability of EFS ± SE and cumulative incidences of events. e Overview of survival and causes of deaths (caused by bleeding, infections or HSCT-related, disease-related deaths or others). Bleeding and infections have been evaluated in patients during initial disease and salvage treatment. Data shown as probability of OS ± SE and cumulative incidences of deaths. *Shorter interval for sufficient follow-up. n = 295 (1987–1992); n = 533 (1993–1998); n = 497 (1999–2004); n = 477 (2005–2010); n = 138 (2011–2012)

Fig. 2

First-line therapy. a Causes of early death from 1987–2012. Data shown as cumulative incidences of deaths ± SE. Gray’s method was used to compare cumulative incidences, shown as p value. Death due to other reasons or inconclusive causes of deaths includes three patients who have been delayed in diagnosis and did not receive treatment in time. n = 295 (1987–1992); n = 533 (1993–1998); n = 497 (1999–2004); n = 477 (2005–2010). b Causes of death during CR from 1987–2012. Data shown as cumulative incidences of deaths ± SE. Gray’s method was used to compare cumulative incidences. n = 295 (1987–1992); n = 533 (1993–1998); n = 497 (1999–2004); n = 477 (2005–2010). c Development of survival after allogeneic HSCT in first CR from 1987–2010. Analysis of survival (pOS and pEFS ± SE) after HSCT in 6-year periods (1987–1992: n = 18; 1993–1998: n = 52; 1999–2004: n = 69; 2005–2010: n = 43). Kaplan–Meier curves of EFS and OS of distinct subgroups were compared using the log-rank test, shown as p value. d Development of survival of all patients without allogeneic HSCT in first CR from 1987–2010. Analysis of survival (pOS and pEFS ± SE) in 6-year periods (1987–1992: n = 277; 1993–1998: n = 481; 1999–2004: n = 428; 2005–2010: n = 434). Kaplan–Meier curves of EFS and OS of distinct subgroups were compared using the log-rank test, shown as p value

Fig. 3

Non-response and relapse. a Protocol flow chart of trials AML-BFM REZ 91, AML-BFM REZ 93, AML-BFM REZ 97, and Relapsed AML 2001/01. HAM high-dose cytarabine, mitoxantrone, SCT stem cell transplantation, MITOX mitoxantrone, E etoposid, A cytarabine (intermediate dose), Dx liposomal daunorubicine, FLAG fludarabine, cytarabine, G-CSF, Consol. consolidation consisting of thioguanine and low-dose cytarabine for 6 weeks. ⁺Low patient recruitment due to alternative regimens, no treatment or palliative care. *Autologous SCT if no suitable matched allogeneic donor available in late relapses. Analysis of pEFS ± SE after relapse (b) and their cumulative incidences of deaths ± SE and pOS ± SE from 1987–2010 (d). n = 97 (1987–1992); n = 159 (1993–1998); n = 172 (1999–2004); n = 156 (2005–2010). Analysis of pEFS ± SE after non-response (c) and their cumulative incidences of deaths ± SE and pOS ± SE (e). n = 48 (1987–1992); n = 56 (1993–1998); n = 46 (1999–2004); n = 47 (2005–2010). Causes of deaths are evaluated for bleeding, infections, SCT-related, disease-related, or others. b–e Kaplan–Meier curves of EFS and OS of distinct subgroups were compared using the log-rank test, shown as p value. BFM Berlin Frankfurt Münster, CNS central nervous system

Fig. 4

Allogeneic HSCT after relapse or non-response. Development of survival after HSCT from 1987–2010. a Analysis of pOS and pEFS ± SE after allogeneic HSCT in 6-year periods in patients with relapse or non-response. (1987–1992: n = 30; 1993–1998: n = 92; 1999–2004: n = 129; 2005–2010: n = 156). b Analysis of 5-year pOS ± SE in patients with early (<1 year of date of diagnosis) or late relapse (>1 year of date of diagnosis) with (+) and without (−) allogeneic HSCT. Early relapse with HSCT n = 5 (1987–1992); n = 31 (1993–1998); n = 46 (1999–2004); n = 63 (2005–2010), early relapse without HSCT n = 35 (1987–1992); n = 52 (1993–1998); n = 43 (1999–2004); n = 18 (2005–2010). Late relapse with HSCT n = 16 (1987–1992); n = 41 (1993–1998); n = 55 (1999–2004); n = 53 (2005–2010), late relapse without HSCT n = 40 (1987–1992); n = 28 (1993–1998); n = 20 (1999–2004); n = 20 (2005–2010). c Analysis of 5-year pOS ± SE after allogeneic HSCT in relapsed AML of standard risk or high-risk AML. Data shown in two time periods ((1987–1998) and (1999–2010)). Standard risk group indicates FAB M1/2 with Auer rods, FAB M4 with atypical eosinophils (M4Eo) and/or favorable cytogenetics, such as t(8;21) and/or AML1-ETO and inv(16) or t(16;16) and/or CBFB/MYH1, if there was no persistence of BM blasts (≥5%) on day. All others were classified as high-risk group. a–c EFS and OS curves of distinct subgroups were compared using the log-rank test, shown as p value

Fig. 5

Pediatric AML from 1987–2010. Fractions of pediatric patients are classified regarding their status quo (alive in relapse, alive in first CR, alive after non-response, death after relapse or non-response and other deaths such as early death or death in CR) from 1987–2010. n = 295 (1987–1992); n = 533 (1993–1998); n = 497 (1999–2004); n = 477 (2005–2010)