1 INTRODUCTION

Acute myeloid leukemia (AML) comprises 15% to 20% of all pediatric leukemias. AML is a life-threatening disease and accounts disproportionately for about one-third of the deaths from childhood leukemia. After a peak in infants, the incidence of AML declines, reaching a minimum in 1 to 9 year old children, which is followed by a gradual increase through adolescence and (young) adulthood.1, 2 The prognosis of pediatric AML patients has improved in the past decades with 5-year survival rates currently about 70% to 75% in Europe and the United States.2-6 The favorable trend in survival of pediatric AML has been attributed to better supportive care, optimization of risk stratification, intensification of chemotherapy, improvements in stem cell transplantation (SCT) and higher salvage rates at relapse.2, 7, 8

Adolescent and young adult (AYA) cancer patients are often reported to have an inferior prognosis compared to children with the same disease. Although the higher rate of survival improvement among AYAs since the early 1990s narrowed the survival gap in Europe, AYAs with cancer still had a worse prognosis than pediatric patients in the mid-2000s. In the same period, survival of AYAs with cancer in the United States increased at a similar rate as that of children and remained to lag behind.9 Age at diagnosis is an important prognostic factor for AML and is inversely associated with the probability of survival.1

So far, population-based survival of children and AYAs diagnosed with AML has been investigated in a limited number of studies from Europe and the United States.10-14 In general, AYAs experienced lower long-term survival compared to pediatric patients; however, the observed differences did not always reach statistical significance, perhaps because of small sample sizes. Moreover, results were generally not specified by cytogenetic risk group, therapy or site of treatment.

Long-term survival of children and AYAs with AML in the Netherlands has not been compared yet, despite the availability of high-quality data collected by the Netherlands Cancer Registry (NCR). According to a recent study focusing on Dutch pediatric AML patients, 5-year overall survival for the period 2010 to 2015 was 84% for younger children (1-9 years) and 66% for older children and adolescents (10-17 years).15 So, increasing age was already associated with worse survival outcome within the pediatric AML population.

In the present study, we investigated long-term survival of children and AYAs (0-39 years) diagnosed with AML in the Netherlands between 1990 and 2015 using population-based, nationwide data from the NCR including information on therapy and site of primary treatment. Our primary objective was to determine whether AYA AML patients (18-39 years) also have a worse prognosis than pediatric AML patients (0-17 years) in the Netherlands and if differences in prognosis have changed over time due to improved survival.

2 PATIENTS AND METHODS 2.1 Data collection

All patients aged 0 to 39 years who were diagnosed with AML in the Netherlands between 1990 and 2015 were identified using the NCR. The NCR is a population-based cancer registry with national coverage since 1989 and has an estimated completeness of at least 96%.16, 17 The NCR receives information concerning all incident cancers diagnosed in the Netherlands from the Nationwide Network and Registry of Histopathology and Cytopathology (PALGA) and the National Registry of Hospital Discharges. After notification, trained registration clerks of the NCR collect data on patient, tumor and treatment characteristics from medical records. Annual linkage with the nationwide Personal Records Database (BRP) is performed to check vital status (last linkage: 1 February 2020).

AML was defined based on International Classification of Childhood Cancer, third edition (ICCC-3), diagnostic group Ib “Acute myeloid leukemias”. AML diagnoses were selected from the NCR using International Classification of Diseases for Oncology, third edition (ICD-O-3), morphology codes 9840, 9841 (ICD-O-2), 9861, 9864 (ICD-O-2), 9865 to 9867, 9869 to 9874, 9877 to 9879, 9891, 9895 to 9898, 9910 to 9912, 9920, 9930, 9931 and 9932 (ICD-O-2). Since 2001, core-binding factor (CBF) leukemia (t[8;21][q22;q22], inv[16][p13.1;q22] or t[16;16][p13.1;q22]) has consistently been registered in the NCR based on morphology code. To obtain information regarding CBF leukemia for children diagnosed during the entire study period, linkage was performed with the database of the Dutch Childhood Oncology Group (DCOG), which includes all patients treated at pediatric oncology centers in the Netherlands.15 Cytogenetic testing for CBF leukemia has been performed more structurally since 2001. Acute promyelocytic leukemia (APL, ICD-O-3 code 9866) and myeloid leukemia associated with Down syndrome (ML-DS, ICD-O-3 code 9898) have a much better prognosis than the other AML subtypes and are treated differently. Therefore, APL patients (N = 202) were analyzed separately. ML-DS patients (N = 28) were excluded from the relative survival, excess mortality and early mortality analyses, because ML-DS was not diagnosed in AYAs. Depending on their age at diagnosis, patients were classified as either children (0-17 years) or AYAs (18-39 years). Patients were followed from their date of diagnosis until death, emigration, loss to follow-up or 1 February 2020 (end of follow-up), whichever occurred first.

2.2 Statistical analyses

Descriptive statistics were used to assess characteristics of the study population. Statistical significance of differences between children and AYAs was tested by Pearson's Χ2 tests or Fisher's exact tests (N ≤ 5 in one or more categories) for categorical variables.

Five- and 10-year relative survival rates were estimated utilizing the traditional cohort approach by applying the strs procedure in Stata18 using 6-month intervals during the first year of follow-up and annual intervals thereafter. Relative survival estimates disease-specific survival by correcting for mortality due to competing causes and is calculated by dividing the observed patient survival by the expected survival of a comparable cohort from the general population.18 Expected probabilities of survival were determined by the Ederer II method19 using Dutch population life tables stratified by age, sex and calendar year, which were obtained from Statistics Netherlands (CBS). Linear trends in relative survival over the diagnostic periods (1990-1999, 2000-2009 and 2010-2015) were assessed by generalized linear models (GLMs) using a Poisson assumption for the observed number of deaths.18 GLMs were also utilized to investigate the relation between age at diagnosis and excess mortality, which is the mortality analogue of relative survival. To adjust for potential confounding, multivariable models were run including sex, diagnostic period, SCT (only for AML) and site of primary treatment. Patients were classified as being treated at an academic hospital when they had received chemotherapy or a SCT at such type of hospital. All GLMs were adjusted for follow-up time in years. The 1 to 9 year age group was used as reference when modeling excess mortality for more detailed age categories, because infants (0 years) with AML have a clearly worse prognosis than older children and the main aim of our study was to compare survival outcomes of children and AYAs. Patients who died on the day of diagnosis were included in the survival models with a follow-up time of 1 day.

Early mortality (death within 30 days of diagnosis) was examined as a secondary outcome to evaluate whether the potentially worse long-term survival of AYAs compared to children could be the result of higher mortality among AYAs shortly after diagnosis. Patients who died on the day of diagnosis were therefore included in the early mortality analyses. Logistic regression analyses were performed to investigate univariable and multivariable associations of age at diagnosis with early mortality. Multivariable analyses were adjusted for sex, diagnostic period and site of primary treatment. Furthermore, diagnostic period was entered as a continuous term into logistic regression models to obtain P-values for linear trends in early mortality over time. Patients diagnosed by autopsy were excluded from the relative survival, excess mortality and early mortality analyses (AML: N = 7, 0.4%; APL: N = 0). Death certificate only (DCO) cases are not included in the NCR as there is no linkage with the cause-of-death registry at an individual basis due to legislation.

CBF leukemia has more consistently been tested and registered since 2001. To evaluate the potentially confounding effect of this favorable risk subtype on the excess mortality estimates, GLMs with and without adjustment for CBF leukemia (no vs yes) were compared for the period 2001 to 2015. Furthermore, relative survival was estimated in children and AYAs diagnosed with CBF leukemia after 2001.

Statistical analyses were executed using Stata 16 (StataCorp LLC, College Station, Texas). Two-sided P-values <.05 were considered statistically significant.

3 RESULTS 3.1 Characteristics study population

In the period 1990 to 2015, 2058 patients aged below 40 years were diagnosed with AML in the Netherlands, of whom 675 were children (0-17 years) and 1383 were AYAs (18-39 years). Patient characteristics are presented in Table 1. The median age at diagnosis was 6 years in children and 31 years in AYAs. Pediatric AML patients were more commonly boys compared to girls, whereas both sexes were fairly equally represented among AYAs. With respect to the AML subtypes, APL occurred more frequently in AYAs (12% vs 5%), whereas the opposite pattern was observed for ML-DS (0% vs 4%) and CBF leukemia (6% vs 14%). The proportion of patients treated at a nonacademic hospital was almost three times higher among AYAs compared to children (21% vs 8%). Of the patients who were treated at a nonacademic hospital, 82% received primary treatment at a teaching hospital. Furthermore, AYAs were nearly twice more likely to receive SCT than children (43% vs 23%). Patients with AML (excluding APL and ML-DS) and APL were further analyzed separately. After exclusion of diagnoses based on autopsy, 1821 AML patients and 202 APL patients could be included in the analyses. ML-DS patients (N = 28) were excluded.

TABLE 1. Characteristics of children (0-17 years) and AYAs (18-39 years) diagnosed with AML in the Netherlands between 1990 and 2015 Characteristics Total Children (0-17 years) AYAs (18-39 years) N % N % N % P (χ2)a Overall 2058 675 1383 Median age at diagnosis in years, IQR 26 21 6 12 31 11 Sex .001 Male 1040 50.5 377 55.9 663 47.9 Female 1018 49.5 298 44.2 720 52.1 Period of diagnosis .04 1990-1999 833 40.5 247 36.6 586 42.4 2000-2009 769 37.4 270 40.0 499 36.1 2010-2015 456 22.2 158 23.4 298 21.6 Subtype <.001 APL 202 9.8 34 5.0 168 12.2 ML-DS 28 1.4 28 4.2 0 0.0 Myeloid sarcoma 21 1.0 9 1.3 12 0.9 CBF leukemia: t(8;21) or inv(16)/t(16;16)b 174 8.5 95 14.1 79 5.7 AML other 1633 79.4 509 75.4 1124 81.3 Site of primary treatment <.001 Nonacademic hospital 346 16.8 51 7.6 295 21.4 Academic hospital 1711 83.2 624 92.4 1087 78.7 Therapy Chemo 1946 94.7 645 95.8 1301 94.1 .11 SCT 744 36.2 155 23.0 589 42.6 <.001 Notes: The % missing values was <1% for all variables included in this table. Statistically significant P values (P < .05) are displayed in bold. Abbreviations: AML, acute myeloid leukemia; APL, acute promyelocytic leukemia; AYAs, adolescents and young adults; CBF, core-binding factor; IQR, interquartile range; ML-DS, myeloid leukemia associated with Down syndrome; SCT, stem cell transplantation. 3.2 Acute myeloid leukemia 3.2.1 Relative survival

The 5-year relative survival of AML patients younger than 40 years in the Netherlands increased by 22 percent-points during 1990 to 2015 from 40% to 62% (Ptrend < .001, Table 2). Similarly, the 10-year relative survival improved from 37% to 47% between 1990 and 2009. Children (N = 612) had a better prognosis than AYAs (N = 1209, Table 2; Figure 1). Overall, 5- and 10-year relative survival were approximately 15 percent-points higher in children compared to AYAs. The rise in 5-year relative survival between 1990 and 2015 was more pronounced among children (+25%, Ptrend < .001) than AYAs (+19%, Ptrend < .001), causing the survival gap to be largest in the latest period. In 2010 to 2015, the 5-year relative survival of children was 74%, which was nearly 20 percent-points higher than the corresponding estimate for AYAs (55%). Furthermore, the survival advantage of children over AYAs generally remained present when considering subgroups based on sex, site of primary treatment and therapy (Table 2). Regarding the AML subtypes, CBF leukemia was more structurally tested and registered as from 2001. Within the group of patients diagnosed with CBF leukemia since 2001 (children: N = 68, AYAs: N = 68), children also had a better 5-year relative survival than AYAs (84% vs 75%), though this difference did not reach statistical significance.

TABLE 2. Five-year and 10-year relative survival of children (0-17 years) and AYAs (18-39 years) diagnosed with AML (excl. APL and ML-DS) in the Netherlands between 1990 and 2015, overall and by subgroup 5-year relative survival (1990-2015) 10-year relative survival (1990-2009)b Total Children (0-17 years) AYAs (18-39 years) Total Children (0-17 years) AYAs (18-39 years) N % SE N % SE N % SE N % SE N % SE N % SE Overall 1821 47.8 1.2 612 57.5 2.0 1209 42.9 1.4 1435 41.8 1.3 473 51.2 2.3 962 37.1 1.6 Sex Male 935 46.9 1.6 341 55.5 2.7 594 42.0 2.0 738 40.4 1.8 270 48.3 3.0 468 35.7 2.2 Female 886 48.8 1.7 271 60.0 3.0 615 43.8 2.0 697 43.3 1.9 203 55.1 3.5 494 38.5 2.2 Period of diagnosis 1990-1999 751 39.5 1.8 229 48.8 3.3 522 35.4 2.1 751 36.9 1.8 229 47.1 3.3 522 32.4 2.1 2000-2009 684 49.3 1.9 244 56.2 3.2 440 45.4 2.4 684 47.1 1.9 244 55.0 3.2 440 42.7 2.4 2010-2015 386 61.6 2.5 139 74.1 3.7 247 54.5 3.2 NA NA NA NA NA NA NA NA NA Ptrend <.001 <.001 <.001 <.001 .04 .001 Site of primary treatment Nonacademic hospital 287 35.0 2.8 44 36.4 7.3 243 34.7 3.1 249 29.4 2.9 43 37.3 7.4 206 27.7 3.1 Academic hospital 1534 50.2 1.3 568 59.1 2.1 966 45.0 1.6 1186 44.4 1.5 430 52.6 2.4 756 39.7 1.8 Therapy Chemo 1731 49.7 1.2 587 59.3 2.0 1144 44.8 1.5 1357 43.6 1.4 451 53.0 2.4 906