Induction mortality and resource utilization in children treated for acute myeloid leukemia at free-standing pediatric hospitals in the United States

Authors

  • Marko Kavcic MD,

    Corresponding author
    1. Division of Oncology, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    • The Children's Hospital of Philadelphia, 3501 Civic Center Boulevard, Philadelphia, PA 19104

    Search for more papers by this author
    • Fax: (267) 425-0113

  • Brian T. Fisher DO, MSCE,

    1. Division of Infectious Diseases, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    2. Center for Pediatric Clinical Effectiveness, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    3. Department of Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Yimei Li PhD,

    1. Division of Oncology, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Alix E. Seif MD, MPH,

    1. Division of Oncology, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    2. Department of Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Kari Torp,

    1. Division of Oncology, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Dana M. Walker MD, MSCE,

    1. Division of Oncology, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    2. Department of Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Yuan-Shung Huang MS,

    1. Department of Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Grace E. Lee MD,

    1. Division of Infectious Diseases, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Sarah K. Tasian MD,

    1. Division of Oncology, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Marijana Vujkovic PhD,

    1. Division of Oncology, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Rochelle Bagatell MD,

    1. Division of Oncology, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    2. Department of Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Richard Aplenc MD, PhD

    1. Division of Oncology, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    2. Center for Pediatric Clinical Effectiveness, The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    3. Department of Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    4. Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author

Abstract

BACKGROUND:

Clinical trials in pediatric acute myeloid leukemia (AML) determine induction regimen standards. However, these studies lack the data necessary to evaluate mortality trends over time and differences in resource utilization between induction regimens. Moreover, these trials likely underreport the clinical toxicities experienced by patients.

METHODS:

The Pediatric Health Information System database was used to identify children treated for presumed de novo AML between 1999 and 2010. Induction mortality, risk factors for induction mortality, and resource utilization by induction regimen were estimated using standard frequentist statistics, logistic regression, and Poisson regression, respectively.

RESULTS:

A total of 1686 patients were identified with an overall induction case fatality rate of 5.4% that decreased from 9.8% in 2003 to 2.1% in 2009 (P = .0023). The case fatality rate was 9.0% in the intensively timed DCTER (dexamethasone, cytarabine, thioguanine, etoposide, and rubidomycin [daunomycin]/idarubicin) induction and 3.8% for ADE (cytarabine, daunomycin, and etoposide) induction (adjusted odds ratio = 2.2, 95% confidence interval = 1.1-4.5). Patients treated with intensively timed DCTER regimens had significantly greater antibiotic, red cell/platelet transfusion, analgesic, vasopressor, renal replacement therapy, and radiographic resource utilization than patients treated with ADE regimens. Resource utilization was substantially higher than reported in published pediatric AML clinical trials.

CONCLUSIONS:

Induction mortality for children with AML decreased significantly as ADE use increased. In addition to higher associated mortality, intensively timed DCTER regimens had a correspondingly higher use of health care resources. Using resource utilization data as a proxy for adverse events, adverse event rates reported on clinical trials substantially underestimated the clinical toxicities of all pediatric AML induction regimens. Cancer 2013. © 2013 American Cancer Society.

INTRODUCTION

Outcomes in pediatric acute myeloid leukemia (AML) have significantly improved over the past decade with remission induction and overall survival rates approaching 90% and 65%, respectively.1 Cooperative group clinical trials have played a critical role in these improved outcomes and in the reduced treatment-related mortality (TRM).2-6 In the United States during the 1990s, these trials focused on intensively timed DCTER (dexamethasone, cytarabine, thioguanine, etoposide, and daunomycin [rubidomycin; rubomycin]/idarubicin)-based regimens. Starting in 2003, the trials shifted to ADE (cytarabine, daunomycin, and etoposide)-based regimens. These clinical trial data have been used to estimate TRM,7 and to evaluate the incidence and risk factors for infectious complications,8 the impact of body mass index,9 and the impact of race/ethnicity10, 11 on overall mortality and TRM.

Although these studies generated important information on induction outcomes, they lack the data needed to compare in temporal trends in regimen use and regimen-related mortality and resource utilization. In addition, cooperative group trials depend on ascertainment and reporting of toxicities by individual centers. Although this reporting mechanism is considered a “gold standard,” recent studies report that clinical trials may underestimate clinically important treatment-related toxicities.12, 13 Thus, published adverse event rates may not accurately reflect the actual clinical experience of patients treated on clinical trials. These limitations have led to a recent call for population-based registry studies in AML.14

Unfortunately, population-based registries of pediatric oncology patients are not currently available in the United States. However, administrative and billing data can be used to establish large cohorts of patients in which to study these important endpoints.15 In order to compare induction mortality and resource utilization by induction regimen, we assembled a nationally representative retrospective cohort of children treated for AML over a 10-year period at 39 free-standing pediatric hospitals contributing to the Pediatric Health Information System (PHIS) database. PHIS data has been used in more than 200 peer-reviewed publications including studies of appendicitis management,16 variation in pediatric antibiotic usage,17 renal failure in pediatric acute myeloid leukemia,18 and end-of-life care in children with cancer.19 PHIS data include 85% of all hospitalizations in free-standing dedicated children's hospitals in the United States registered by the National Association of Children's Hospitals (D. Bertoch, personal written communication, December 8, 2010), including approximately 40% of patients treated on the recently completed Children's Oncology Group (COG) phase 3 trial (AAML0531) for children with newly diagnosed AML.20

We hypothesized that PHIS data could be used to define temporal trends of chemotherapy regimen use and induction mortality and to compare resource utilization between induction regimens. We further hypothesized that intensively timed DCTER induction regimens would be more resource-intensive than the ADE regimen and that PHIS resource utilization data would allow quantification of clinically relevant adverse events (AEs) not reported by the clinical trials.

MATERIALS AND METHODS

Human Subjects

The study was approved by the Children's Hospital Association (CHA) and received exemption status by the Institutional Review Board at the Children's Hospital of Philadelphia.

Data Source

PHIS is a comparative pediatric administrative database including inpatient data from 42 not-for-profit, tertiary children's hospitals affiliated with the CHA (Overland Park, Kan).15 PHIS data are derived from 2 primary data sources within the participating hospitals. The hospital's medical record system provides patient identification, demographics, dates of service, discharge disposition, and up to 41 diagnosis and procedure codes from the International Classification of Diseases, Ninth Revision (ICD-9). The hospital's billing system provides billed resource utilization data for every patient for every hospital day of service. These resource utilization data include all billed pharmaceuticals, laboratory tests (without results), imaging procedures (without results), and supplies. Each patient's vital status at the time of the hospital discharge is also recorded. Each patient is assigned a unique identifier in the PHIS database that is preserved for all admissions. Therefore, patients can be followed across inpatient admissions. Oversight of PHIS data quality is a joint effort between the CHA (data management center), Thomson Reuters Healthcare (data processing partner), and participating hospitals.

Study Cohort Assembly

Patients with hospital discharge dates in the PHIS database between January 1, 1999, and March 31, 2010, were included in the study cohort. Children were included in the study cohort if their index hospitalization was assigned an ICD-9 code (either as a primary or secondary diagnosis) for any type of myeloid or unspecified leukemia (205.xx: myeloid leukemia, 206.xx: monocytic leukemia, 207.xx: other specified leukemia, and 208.xx: leukemia of unspecified cell type).21

Patients were excluded if they had been assigned an ICD-9 discharge diagnosis code consistent with another malignancy or had undergone hematopoietic stem cell transplantation within 60 days of index admission. Induction chemotherapy billing data for the remaining patients were subsequently manually reviewed and compared with published regimens2, 22, 23 by a pediatric oncologist (M.K.).15, 21 Those patients not receiving chemotherapy consistent with published AML induction regimens were excluded. Induction regimens were categorized as 7+3 (cytarabine and daunorubicin), ADE (cytarabine, daunorubicin, and etoposide) with or without gemtuzumab ozogamicin, DAT (daunorubicin, cytarabine, thioguanine), intensively timed DCTER (2 intensively timed cycles of dexamethasone, cytarabine, thioguanine, etoposide, and rubidomycin [daunomycin] or idarubicin given 10 days apart), or other (standard timing of DCTER and other regimens). The induction period was defined from initiation to completion of 2 courses of chemotherapy.1-3, 22

Study Variables

Demographics

Demographic data (age, sex, and race) and insurance status were defined at the first admission with chemotherapy. Age was collected in years, and sex, age, race (white, black, Asian, Native American, other, or missing), and insurance status (private, government, self-pay, other, or unknown) were each collected as categorical variables. Because ethnicity data were missing for a substantial number of patients, no analyses by ethnicity were conducted.

Hospital Resources Utilized

The frequency of inpatient resource use was determined for each patient based on daily billing data. Utilization frequencies were determined for the following resources: complete blood counts, blood cultures, antimicrobial agents, blood products, opioid analgesics, antihypertensive medications, vasopressors, diuretics, antiemetics, supplemental oxygen, mechanical ventilation, extracorporeal membrane oxygenation, radiology imaging, parenteral nutrition, nasogastric tube placement, and dialysis. Each resource variable was first dichotomized (exposure or no exposure) for each inpatient day, then summarized as the total number of exposure days during the induction period for each patient.

Hospital Discharge Status

PHIS disposition data document the conclusion of each hospitalization classified as: discharged to home, discharged to another care facility, or inpatient death. Using these disposition data, the inpatient case fatality rate during the induction period was calculated for each induction regimen.

Statistical Analyses

Baseline and demographic characteristics were summarized by standard descriptive statistics. Chi-square tests were used to evaluate for variation in the distribution of each demographic characteristic among the 5 regimen groups. Logistic regression analyses were performed to test the piecewise linear trend in the mortality rate over time.

Univariate logistic regression was used to evaluate unadjusted effects of demographic characteristics and induction regimen on inpatient induction mortality risk. Multivariable logistic regression was used to determine the adjusted association of induction regimens with inpatient mortality controlling for patient characteristics and time trends. Results were summarized as unadjusted and adjusted odds ratios (ORs) with 95% confidence intervals (CIs). Because both the exposure of interest (regimen use) and outcome (inpatient mortality) clustered by hospital, adjustment for confounding by hospital was performed. The effects of induction regimen were decomposed to patient level and hospital level24, 25 by including percent of each regimen use in a hospital as hospital level covariates in the multivariable regression. In this adjustment, for example, the patient level effect measures the effect of receiving DCTER versus ADE once a child is admitted to a hospital, and the hospital level effect measures the effect of transferring a given child from a hospital with lower to higher percent of DCTER use.

In the resource utilization analysis, resource utilization days per 1000 inpatient days were reported. Poisson regressions were used to compare the rates of resource utilization between different regimen groups with the resource days as the outcome and inpatient days as the offset. To adjust for potential overdispersion in Poisson regressions, the Pearson scale adjustment was applied.

RESULTS

Between January 1, 1999, and March 31, 2010, 1686 patients with 132,330 inpatient days (362.2 years) of observation time and presumed de novo AML were identified at 39 of 42 PHIS institutions.21 Table 1 lists patient characteristics by induction regimen. The median age of the patients was 7.7 years. A slight male predominance was observed (53%), most patients were white (70%), and more patients had government (40%) than private (34%) insurance at their index admission. Age distribution differed significantly among regimens with nearly half (46.5%) of patients treated with DAT being between 1 and 3 years of age (P < .0001). Values of the other baseline characteristics (sex, race, insurance status) were equally distributed among induction regimens.

Table 1. Baseline Characteristics of Patients (N = 1686) by Induction Regimen
CharacteristicAll Patients N = 1686ADE n = 931DAT n = 243DCTER n = 3337+3 n = 84Othera n = 95Pb
 N%n%n%n%n%n% 
  • a

    “Other” includes MA (mitozantrone and cytarabine) with or without gemtuzumab ozogamicin, MAE (mitozantrone, cytarabine, and etoposide), and standard timing of DCTER.

  • b

    P value is from chi-square test to compare the demographic distribution among 5 regimen groups.

  • Abbreviations: ADE, cytarabine, daunomycin, and etoposide with or without gemtuzumab ozogamicin; DAT, daunorubicin, cytarabine, thioguanine; DCTER, 2 intensively timed cycles of dexamethasone, cytarabine, thioguanine, etoposide, and rubidomycin (daunomycin) or idarubicin given 10 days apart; 7+3, cytarabine- and daunorubicin-based induction regimen.

Age at start of chemotherapy, y            <.0001
 <117510.49510.2239.53610.8910.71212.6 
 ≥1 and <339423.417819.111346.56720.11113.12526.3 
 ≥3 and <1039423.421322.95121.09027.02226.21818.9 
 ≥10 and <1543525.826728.73213.29027.02529.82122.2 
 ≥15 and <1928817.117819.1249.95015.01720.21920.0 
Sex            .95
 Male89853.349653.312451.018054.04654.95254.7 
 Female78846.743546.711949.015346.03845.24345.2 
Race            .12
 White127469.664369.117873.322367.06467.46678.6 
 Black22313.212413.33514.44714.188.4910.7 
 Asian/Pacific Islander603.6343.772.9113.355.333.6 
 American Indian130.880.90041.211.100 
 Other1408.3609.7145.8247.21010.522.4 
 Unknown764.5323.493.7247.277.444.8 
Insurance at start of chemotherapy            .51
 Private57834.332534.97129.212537.52930.52833.3 
 Government68040.338441.39438.712937.84244.23440.5 
 Self-pay382.3192.062.582.411.144.8 
 Other38823.020221.77129.27422.22324.21821.4 
 Unknown20.110.110.4000000 

Figure 1 illustrates the variation in induction regimens and inpatient case fatality rate through the study period. Although most children (75%) were treated with DCTER or ADE regimens, substantial variation over time existed. In 2004, use of DAT and DCTER regimens declined. Simultaneously, ADE chemotherapy became more widely utilized and by 2009 comprised 87% of induction regimens. The case fatality rate did not increase significantly between 1999 and 2003 (6.7% versus 9.8%; P = .4813, test for trend), but substantially decreased after 2003 to a rate of 2.1% in 2009 (P = .002, test for trend).

Figure 1.

Induction regimen use and case fatality rate with 95% confidence intervals are shown by study year. Abbreviations: 7+3, cytarabine- and daunorubicin-based induction regimen; ADE, cytarabine, daunomycin, and etoposide with or without gemtuzumab ozogamicin; DAT, daunorubicin, cytarabine, thioguanine; DCTER, 2 intensively timed cycles of dexamethasone, cytarabine, thioguanine, etoposide, and rubidomycin (daunomycin) or idarubicin given 10 days apart.

Unadjusted and adjusted ORs for induction mortality by induction regimen are shown in Table 2. Case fatality rate varied by induction regimen from 3.7% for DAT to 9.0% for DCTER. In the fully adjusted analysis, age was a significant predictor of induction mortality, with infants and adolescents having higher mortality risk (OR = 4.3; 95% CI = 2.2-8.7, and OR = 3.5; 95% CI = 1.8-6.0, respectively). Black patients did not have a higher likelihood of death than white patients (adjusted OR = 1.4; 95% CI = 0.7-2.7). Patients in race group “Other” (Asians/Pacific Islanders, American Indians, other, and unknown race) were twice as likely to die in induction than white patients (adjusted OR = 2.0; 95% CI = 1.2-3.4). Insurance status at index admission had no effect on mortality risk.

Table 2. Risk Factors for Induction Mortality
FactorUnadjusted ORUnadjusted 95% CIAdjusted ORAdjusted 95% CI
  • Unadjusted odds ratio (OR) derived from univariable logistic regression; adjusted OR derived from multivariable logistic regression including patient demographic characteristics and hospital level regimen use.

  • Abbreviations: ADE, cytarabine, daunomycin, and etoposide with or without gemtuzumab ozogamicin; CI, confidence interval; DAT, daunorubicin, cytarabine, thioguanine; DCTER, 2 intensively timed cycles of dexamethasone, cytarabine, thioguanine, etoposide, and rubidomycin (daunomycin) or idarubicin given 10 days apart; 7+3, cytarabine- and daunorubicin-based induction regimen.

  • *

    P < .05,

  • **

    **P < .01,

  • ***

    P < .001

  • a

    “Other” includes Asian/Pacific Islander, American Indian, other, and unknown.

  • b

    “Other” includes self-pay, other, and unknown.

  • c

    “Other” includes MA (mitozantrone and cytarabine), MA with gemtuzumab ozogamicin, MAE (mitozantrone, cytarabine, and etoposide), and standard timing of DCTER.

Age    
 <13.82***(1.94,7.53)4.33***(2.16,8.72)
 ≥1 and <30.53(0.22,1.25)0.55(0.22,1.32)
 3 and <10ReferenceReferenceReferenceReference
 ≥10 and <150.90(0.44,1.87)0.88(0.42,1.84)
 ≥15 and <192.94**(1.55,5.57)3.54***(1.83,6.84)
Sex    
 MaleReferenceReferenceReferenceReference
 Female1.42(0.93,2.17)1.57*(1.00,2.44)
Race    
 WhiteReferenceReferenceReferenceReference
 Black1.26(0.68,2.35)1.37(0.71,2.66)
 Othera1.76*(1.06,2.91)1.99*(1.15,3.43)
Insurance at first admission    
 PrivateReferenceReferenceReferenceReference
 Government1.17(0.72,1.90)1.17(0.69,1.97)
 Otherb0.90(0.50,1.60)0.82(0.44,1.53)
Year (current vs previous)    
 1999-20031.07(0.88,1.31)1.11(0.89,1.38)
 2004-20100.83**(0.73,0.93)0.87(0.75,1.00)
Induction regimen, patient level    
7+31.97(0.80,4.83)2.24(0.79,6.32)
ADEReferenceReferenceReferenceReference
DAT0.98(0.47,2.08)0.88(0.35,2.23)
DCTER2.53***(1.53,4.20)2.17*(1.06,4.46)
Otherc3.35***(1.64,6.84)2.76*(1.21,6.33)

Table 3 compares induction duration, days of in-hospital stay during induction period, and summary data for resource utilization by induction regimen. For 22 of the 36 resource metrics measured, patients receiving intensively timed DCTER chemotherapy had significantly increased resource utilization compared to those treated with ADE regimens. Notably, patients treated with ADE were hospitalized an average of 16 fewer days, had 50% lower rates of parenteral nutrition and patient-controlled analgesia usage, a 30% lower rate of both blood culture testing and vasopressor infusion administration (norepinephrine, epinephrine, and dobutamine), and a 20% lower rate of blood product (packed red blood cell and platelet) transfusions. Multiple other resources, including overall antibiotic use, abdominal ultrasound, computed tomography imaging of the head, and oral and parenteral analgesics were also significantly decreased in patients receiving ADE induction regimens.

Table 3. Induction Resource Utilization Rate (Days of Resource Exposure per 1000 Hospital Days) by Regimen
Resources MeasuredAll PatientsDCTERADEPa
  • a

    P value of comparison of intensively timed DCTER and ADE regimens.

  • Abbreviations: ADE, cytarabine, daunomycin, and etoposide with or without gemtuzumab ozogamicin; DCTER, 2 intensively timed cycles of dexamethasone, cytarabine, thioguanine, etoposide, and rubidomycin (daunomycin) or idarubicin given 10 days apart.

Mean duration of induction period (days)79.7101.774.4 
Average inpatient stay (days)55.171.755.3<.0001
Complete blood count sampling866.7871.8867.7.61
Blood culture230.4293.5201.4<.0001
Antibiotics    
Total1127.21230.41111.2<.0001
 Broad Gram-positive coverage (vancomycin, linezolid, daptomycin, quinopristin/dalfopristin)317.1340.4316.0.04
 Beta lactam anti-Pseudomonas coverage (ceftazidime, cefepime, piperacillin/tazobactam, ticarcillin/clavulanate)458.1438.2465.1.04
 Carbapenems with anti-Pseudomonas activity (imipenem, meropenem)126.3163.9121.0.0002
 Quinolones27.241.022.7.0001
 Aminoglycosides178.6216.6169.5<.0001
Antifungals    
Total736.5734.6761.1.13
 Amphotericins165.8263.3123.2<.0001
 Echinocandins44.528.160.8.0001
 Azoles526.2443.1577.1<.0001
Antivirals115.3107.5128.0.15
Blood products    
 Packed red blood cells127.3148.4117.6<.0001
 Platelets206.7239.3190.7<.0001
 Fresh frozen plasma15.315.813.9.48
 Cryoprecipitate3.42.93.3.78
Analgesics    
 Nonopioid283.2308.9266.7.0008
 Opioid (all routes)255.4306.2238.5<.0001
 Patient-controlled analgesia35.155.423.6<.0001
Antiemetics474.0413.7518.3<.0001
Parenteral nutrition195.9295.4156.2<.0001
Antihypertensives53.650.955.7.66
Diuretics74.775.965.1.18
Vasopressors    
 Dopamine11.516.48.1.0021
 Other (norepinephrine, epinephrine, dobutamine)18.524.616.7.02
Supplemental oxygen45.372.024.2<.0001
Ventilation27.025.821.8.45
Extracorporeal membrane oxygenation0.60.50.9.76
Dialysis4.98.93.2.02
Imaging    
 Chest x-ray88.083.081.1.78
 Abdominal ultrasound6.27.45.4.009
 Head computed tomography8.69.28.50.61
 Chest computed tomography17.521.317.30.002
 Head magnetic resonance imaging3.23.13.30.68

DISCUSSION

To our knowledge, this is the first study to compare induction mortality and resource utilization using administrative/billing data among induction regimens in children with de novo AML. Although it is a new cohort, it has multiple characteristics demonstrating face validity. First, the age, sex, and racial distributions are consistent with previously published data for pediatric patients with de novo AML.26 Second, the observed induction mortality rates are highly concordant with those published in multiple cooperative group trial analyses.2-4, 22 Finally, the durations of hospitalization for ADE and intensively timed DCTER regimens parallel those reported in data from cooperative group trials.2, 3

As initially hypothesized, patients receiving intensively timed DCTER regimens had a 2-fold increased mortality. Because in-hospital mortality could be confounded by hospital level clustering of regimen and mortality, the proportion of each regimen use at each hospital was included in the multivariable analysis. Even after accounting for hospital clustering and adjusting for other confounding variables, a 2-fold higher risk for induction mortality persisted for patients receiving the intensively timed DCTER regimens. This increase in mortality was reflected in increased resource utilization across a majority of the resource utilization categories examined, suggesting a concordant increase in morbidity. The published data reporting comparable event-free survival rates in intensively timed DCTER and ADE induction regimens,2-5 and our data demonstrating significantly increased induction mortality and resource utilization with intensively timed DCTER therapy support using ADE as the standard induction regimen in children with de novo AML.

Although the clinical trial reports have described higher TRM with intensively timed DCTER regimens, these reports provide very limited data on treatment complications in patients who do not experience TRM. For example, the Children's Cancer Group (CCG) 2961, COG AAML031P1, and Medical Research Council (MRC) AML12 trials did not report pain as a clinically significant AE.2-4 However, per PHIS billing data, patients received opioid medications an average of 1 in 4 hospital days, and the rate of patient-controlled analgesia use in the intensively timed DCTER cohort was more than twice that in the ADE cohort. Similarly, hypotension as an AE is reported only on the AAML031P1 trial with rates of 2% and 3% in induction courses I and II.3 However, PHIS billing data demonstrate that 13% of patients received at least 2 consecutive days of vasopressor support during induction therapy. Such discrepancies occur across almost all resource utilization categories and exemplify how reporting mechanisms on clinical trials substantially and systematically underreport important clinical toxicities.

Although other investigators have raised concerns about AE reporting on clinical trials,13, 27, 28 we believe our analysis is the first to demonstrate such marked underreporting across a wide range of specific toxicities. Work is ongoing to compare AE rates from patients enrolled in COG trials with AE rates estimated from resource utilization metrics. Based on the estimates of Roche et al, the most recent COG pediatric AML trial required 8300 hours of time to report approximately 11,000 adverse events (TA Alonzo, personal written communication, September 20, 2011). Given the large number of AEs per patient and the diminishing resources available for cooperative group clinical trials, the use of day-by-day resource utilization data from administrative/billing data sources may be an untapped resource to complement toxicity ascertainment and consequentially improve its efficiency and accuracy.

Overall yearly inpatient induction case fatality rates dropped 3-fold during the study period. Patients treated from 1999 to 2003 had slightly increasing mortality by study year, but this trend substantially reversed after 2003 when case fatality rates decreased from 9.8% in 2004 to 2.1% in 2009. The etiology of this decline in mortality is multifactorial, but the inversely proportional use of intensively timed DCTER and ADE regimens likely contributes substantially. In addition, a “learning curve” effect for the administration of pediatric AML therapy may also have contributed to this decreased mortality.2

Age and sex were significantly associated with induction mortality. Consistent with prior publications, infants and adolescents each had a 3-fold higher risk for induction mortality in comparison to patients in the reference age group (≥ 3 years and < 10 years).2, 3, 22 Prior reports have not demonstrated increased mortality in female patients; additional work is ongoing to further explore this finding. Although black patients did not have higher induction mortality, “Other” patients, including Asians/Pacific Islanders, American Indians, other, and unknown race, showed a 2-fold higher risk. Racial disparities in pediatric cancer and particularly in childhood AML remain incompletely understood despite multiple investigations.29 Published COG and Surveillance, Epidemiology and End Results (SEER) data have demonstrated decreased overall survival in black patients.3, 10, 30 Single-institution studies have yielded conflicting results with the most recent data demonstrating no difference in outcome.11, 31 However, none of these reported a difference for Asians/Pacific Islanders and American Indians. To address further potential racial/ethnic disparities, we are working to increase the patient cohort size and to evaluate the association between race, resource utilization, and mortality in patients treated with ADE chemotherapy.

This cohort has a large sample size and well-delineated resource utilization information, but PHIS administrative/billing data has limitations. First, the process of cohort assembly may include patients with a diagnosis other than de novo AML.15 However, previous work in acute lymphoblastic leukemia would suggest that the estimated number of such patients is small.15 Second, because deaths occurring at non-PHIS hospitals or in the outpatient setting were not ascertained, our analyses may underestimate the overall induction mortality rate. Because AML induction is delivered primarily in the inpatient setting, this potential bias is also likely modest. In addition, the available resource administration data only describe the services that were billed rather than the actual services provided to the patient. Despite this potential source of error, the relative comparison of resource intensity between regimens should be valid. Moreover, the resource utilization data reflect the treatment burden of each induction regimen on payors of health care. Finally, Hispanic ethnicity data and laboratory/radiology result data were not available for analysis, and therefore analysis of ethnicity and disease-free survival could not be performed. The merging of Children's Oncology Group AML trial data with PHIS20 and the inclusion of clinical test results from a subset of PHIS hospitals32 will help to address these limitations in future studies.

In summary, our analysis is the first comparison of mortality and resource utilization across multiple pediatric AML induction regimens administered in the United States. These data demonstrate a significant decrease in AML induction mortality after 2004 and show that intensively timed DCTER regimens have higher induction mortality and concordantly increased use of health care resources. In addition, the resource utilization data indicate that adverse event rates reported on clinical trials substantially underestimate the actual resources needed to provide de novo AML care. These results provide support for the use of ADE-based regimens as the standard of care for children with newly diagnosed AML in the United States and serve as an impetus for improving adverse event monitoring and reporting on pediatric AML clinical trials. Work is ongoing to characterize further the recent decrease in AML induction mortality and the disparities in outcomes and resource utilization for sex and specific minority groups receiving ADE therapy, as well as to leverage PHIS data to supplement AE reporting data from recently completed and ongoing COG AML clinical trials.20

FUNDING SOURCES

No specific funding was disclosed.

CONFLICT OF INTEREST DISCLOSURES

This study was supported by the US National Institutes of Health grant R01CA165277 to Dr. Aplenc and by The Slovenian Ministry of Science grant J3-4220 to Dr. Kavcic.

Ancillary