Details of therapy are in the ”Methods” section and in Barnard
et al10 and Lange et al.11 GCSF indicates granulocyte
BMI indicates body mass index.
Lange BJ, Gerbing RB, Feusner J, Skolnik J, Sacks N, Smith FO, Alonzo TA. Mortality in Overweight and Underweight Children With Acute Myeloid
Leukemia. JAMA. 2005;293(2):203-211. doi:10.1001/jama.293.2.203
Author Affiliations: Division of Oncology (Dr
Lange), The Children’s Hospital of Philadelphia (Dr Skolnik and Ms Sacks),
Philadelphia, Pa; Children’s Oncology Group, Arcadia, Calif (Mr Gerbing);
Children’s Hospital of Oakland, Oakland, Calif (Dr Feusner); Cincinnati
Children’s Hospital Medical Center, Cincinnati, Ohio (Dr Smith); and
University of Southern California and Children’s Oncology Group, Los
Angeles (Dr Alonzo).
Context Current treatment for acute myeloid leukemia (AML) in children cures
about half the patients. Of the other half, most succumb to leukemia, but
5% to 15% die of treatment-related complications. Overweight children with
AML seem to experience excess life-threatening and fatal toxicity. Nothing
is known about how weight affects outcomes in pediatric AML.
Objective To compare survival rates in children with AML who at diagnosis are
underweight (body mass index [BMI] ≤10th percentile), overweight (BMI ≥95th
percentile), or middleweight (BMI = 11th-94th percentiles).
Design, Setting, and Participants Retrospective review of BMI and survival in 768 children and young adults
aged 1 to 20 years enrolled in Children’s Cancer Group-2961, an international
cooperative group phase 3 trial for previously untreated AML conducted August
30, 1996, through December 4, 2002. Data were collected through January 9,
2004, with a median follow-up of 31 months (range, 0-78 months).
Main Outcome Measures Hazard ratios (HRs) for survival and treatment-related mortality.
Results Eighty-four of 768 patients (10.9%) were underweight and 114 (14.8%)
were overweight. After adjustment for potentially confounding variables of
age, race, leukocyte count, cytogenetics, and bone marrow transplantation,
compared with middleweight patients, underweight patients were less likely
to survive (HR, 1.85; 95% confidence interval [CI], 1.19-2.87; P = .006) and more likely to experience treatment-related
mortality (HR, 2.66; 95% CI, 1.38-5.11; P = .003).
Similarly, overweight patients were less likely to survive (HR, 1.88; 95%
CI, 1.25-2.83; P = .002) and more likely
to have treatment-related mortality (HR, 3.49; 95% CI, 1.99-6.10; P<.001) than middleweight patients. Infections incurred during the
first 2 courses of chemotherapy caused most treatment-related deaths.
Conclusion Treatment-related complications significantly reduce survival in overweight
and underweight children with AML.
Each year in the United States, 500 to 600 individuals younger than
21 years develop acute myeloid leukemia (AML).1 Current
treatment for AML typically consists of 3 or 4 courses of intensive, myelosuppressive
chemotherapy with or without bone marrow transplantation from a histocompatible
family donor. This therapy cures about half the children with AML; of the
other half, most succumb to AML-related causes, but 5% to 15% die from toxic
effects of treatment.2- 4 Factors
that predict treatment failure and death in AML are relatively older age and
higher white blood cell (WBC) count at diagnosis, a slow response to the first
course of chemotherapy,3- 6 and,
in the United States, black race and absence of a histocompatible family member
to donate marrow for transplantation.7,8
Children’s Cancer Group (CCG)-2961 was a phase 3 international
cooperative group trial for pediatric patients with untreated AML. It was
the impression of investigators (J.F., B.J.L.) on this trial that overweight
children and adolescents experienced more toxicity and death than did the
other patients. This observation prompted the following retrospective investigation
of effects of body mass index (BMI) on survival and treatment-related mortality.
CCG-2961 opened on August 30, 1996, and closed on December 4, 2002.
Patients from birth through age 20 years were enrolled after institutional
review board approval of each participating institution and written informed
consent. Patients with Down syndrome, Fanconi anemia, acute promyelocytic
leukemia, acute undifferentiated leukemia, or treatment-related AML were excluded
from the study. The trial accrued 902 patients with de novo AML. Thirty patients
without outcome data and 104 infants younger than 1 year were excluded from
this analysis, leaving 768 patients.
AML was classified according to French-American-British criteria.9 Morphology, histochemistry, and karyotype were centrally
reviewed as described.10,11 Cytogenetic
subsets were classified as normal, favorable [t(8;21); t(9;11) or (inv 16)],
unfavorable [del(7) or 7q-], or standard.3,5,6
Figure 1 illustrates the treatment
plan for CCG-2961. Course 1 consisted of idarubicin, dexamethasone, cytarabine,
thioguanine, and etoposide on days 0 to 3 and daunorubicin plus the last 4
drugs on days 10 to 13. Patients in complete or partial remission were randomized
to course 2 therapy. Complete remission was defined as less than 5% marrow
blasts, with recovery of neutrophils to greater than 1000 × 109/L and platelets to greater than 50 000 × 103/μL,
and partial remission was defined as 5% to less than 30% marrow blasts. Course
2 consisted of a repetition of induction therapy (regimen A) or fludarabine,
idarubicin, and cytarabine (regimen B).12 After
course 2, patients in complete remission and with a histocompatible relative
able to donate marrow were assigned to marrow transplantation; those without
donors were assigned high-dose cytarabine/L-asparaginase chemotherapy.4 Patients in remission after course 3 chemotherapy
were randomized to standard follow-up or to interleukin 2.11 All
patients received central nervous system prophylaxis with 8 doses of intrathecal
cytarabine. Granulocyte colony-stimulating factor was given on day 14 ± 1
of courses 1 and 2 and continued until the absolute neutrophil count (ANC)
was at least 1000 × 109/L.13 All
systemic chemotherapy was given in the hospital.
Data entered onto study forms were abstracted from the medical record.
They included age, sex, ethnicity, height and weight before each phase of
therapy, doses of chemotherapeutic agents, toxicities, infectious complications,
duration of hospitalization, and time to neutrophil recovery to 500 ×
109/L and 1000 × 109/L. Treating physicians or
nurses classified the individual’s race and ethnicity. Drug dosing was
based on weight in kilograms up to age 3 years and by surface area thereafter.
Dose modifications were provided for hyperbilirubinemia and after May 2001
for reduced glomerular filtration or creatinine clearance. There were no dose
modifications for underweight or overweight patients.
Body mass index at diagnosis was calculated as weight in kilograms divided
by the square of the height in meters.14 For
patients older than 2 years, underweight was defined as BMI less than or equal
to the 10th percentile, overweight as BMI greater than or equal to the 95th
percentile, and middleweight as BMI greater than the 10th to less than the
95th percentile (11th-94th percentiles). For patients aged 1 to 2 years, greater
than or equal to the 95th percentile and less than or equal to the 10th percentile
of weight for length were used to define overweight and underweight, respectively.14
The main outcome measures were remission status after courses 1 and
2 of chemotherapy, overall survival, and treatment-related mortality. Survival
was defined as time from registration to death. Treatment-related mortality
was defined as time until death from causes other than AML, censoring for
progressive disease, relapse, and failure to enter remission after 2 courses
of therapy. Patients lost to follow-up were censored at their last known date
of contact. Patients in marrow remission (<5% marrow blasts) but without
recovery of peripheral counts were censored at the end of course 2. Toxicity
grades 3 and 4 were defined by contemporary CCG toxicity grading criteria
with protocol-specific modifications to capture details of anticipated gastrointestinal
and hematopoietic toxicity.
This report analyzes data collected through January 9, 2004, with a
median follow-up of 31 months and a range of 0 to 78 months. To compensate
for the tendency for bad news (ie, deaths and relapses) to be reported sooner
than ongoing follow-up for patients in continuing remission, events such as
deaths and relapses were censored on July 9, 2003, 6 months before data cutoff.
The Kaplan-Meier method was used to calculate survival estimates. Confidence
intervals (CIs) were calculated according to the Greenwood formula. Hazard
ratio (HR) of overweight and underweight patients relative to middleweight
patients was estimated by Cox proportional hazards models. Hazard ratio was
also estimated with multivariate Cox proportional hazards models after adjustment
for standard variables that predict outcomes in AML (black vs other race),
age (<2 years, 2 to ≤ 10 years, and >10 years), cytogenetics,
and WBC count (<50 000 vs ≥50 000 × 109/L), as well
as bone marrow transplantation. Multivariate analysis included patients with
complete covariate data.
Cumulative incidence of neutrophil recovery was estimated by considering
death before recovery as a competing event. Cumulative incidence of completing
course 1 in remission was estimated by considering disease progression and
death before course completion as competing events. Significance of observed
differences in proportions was tested using the χ2 test and
Fisher exact test when data were sparse. For continuous data, the Mann-Whitney
test compared medians of distributions. P≤.05
was set as the threshold for significance.
Of 768 patients at diagnosis, 114 (14.8%) were overweight and 84 (10.9%)
underweight. Table 1 lists demographic
features of overweight, middleweight, and underweight patients, characteristics
of their AML, and their treatment assignments. The 3 BMI groups did not differ
significantly in distribution by sex, ethnicity or race, or the proportion
of patients with incomplete data. Compared with middleweight patients, overweight
patients had higher leukocyte counts (P = .001),
were marginally more likely to have unfavorable marrow cytogenetics (P = .01), were less likely to have a related
marrow donor for marrow transplantation (P = .05),
but were equally likely to actually undergo transplantation (9/10 vs 91/106).
Underweight patients were younger (P = .04)
and more commonly had unfavorable marrow cytogenetics (P = .048).
Table 2 shows the response to
therapy for the 3 BMI groups. In course 1, there were no differences in the
early response rate as measured by the day 14 marrow blast percentage or in
treatment failure rate. There was a trend for a reduced remission rate and
an increased death rate among overweight patients, whereas the underweight
patients showed a definite reduction in remission rate and increase in death
rate. At the end of course 2, compared with middleweight patients, overweight
patients were significantly more likely to die (17% vs 5%, P = .001). The actuarial survival after study enrollment
of underweight and overweight patients was inferior to that of middleweight
patients (Figure 2).
After course 1, the univariate HR for treatment-related mortality was
significantly increased in the overweight patients (HR, 2.12; 95% CI, 1.37-3.28; P = .001) and the underweight patients (HR, 1.80;
95% CI, 1.06-3.06; P = .03) compared with
the middleweight patients (Table 3).
Survival from study entry was reduced in the overweight patients (HR, 1.47;
95% CI, 1.09-1.98; P = .01) and the underweight
patients (HR, 1.42; 95% CI, 1.00-2.03; P = .05).
From the end of course 2, there were no differences in survival among the
3 groups or in relapse rates or treatment-related mortality.
Table 3 shows the multivariate
analysis of HR for survival and treatment-related mortality adjusted for age,
race, WBC count, cytogenetics, and allogeneic bone marrow transplantation
for the patients with complete data for the 5 variables. After adjustment,
overweight and underweight groups were still less likely to survive than middleweight
patients, and HR for treatment-related mortality was even higher than in univariate
Table 4 lists deaths according
to when they occurred and attribution of cause. Compared with middleweight
patients, overweight and underweight patients were more likely to die before
or during their first remission (P = .002
and P = .047, respectively). In all groups,
infection was the most common cause of death before or during remission. After
recurrence of AML, the most common cause of death was AML itself in all 3
Excessive treatment-related mortality suggests that overweight and underweight
patients could be receiving too much chemotherapy. To address this issue,
first we examined the doses of protocol therapy actually received. Then we
compared toxicity grades 3 and 4, time to neutrophil recovery, and duration
of course as indirect measures of drug effect on normal marrow. Of 768 patients,
10 received less than or equal to 90% of dosing of course 1 therapy according
to square meters: in 6 (1 underweight, 1 overweight, and 4 middleweight) of
these 10 patients, doses were reduced according to protocol guidelines for
hyperbilirubinemia. Four overweight patients received reduced doses calculated
to fall between their actual weight and their ideal body weight. Overweight
patients were significantly more likely to have a dose reduction than middleweight
patients: 4.8% vs 0.7% (P = .006).
Toxicity grades 3 and 4 were assessed from a menu of 45 clinical and
laboratory parameters. Compared with middleweight patients, more overweight
patients experienced grade 3 or 4 abdominal pain (P = .05),
systolic hypertension (P = .02), pulmonary
function abnormalities (P = .03), and coagulopathy
(P<.001), whereas more underweight patients experienced
grade 3 or 4 elevations of hepatic enzymes (alanine aminotransferase, P = .01; and aspartate aminotransferase, P = .04).
Death from infection increases in direct proportion to the magnitude
and duration of neutropenia.15 Hematologic
toxicity was probably the most relevant complication for this study. Important
landmarks for assessing neutropenia are time to an ANC of 500 × 109/L (the time when empirical antibiotics are typically discontinued)
and time to an ANC of 1000 × 109/L (the time when the patient
may be able to begin the next chemotherapy course). Within 5 weeks of the
start of therapy, 79.5% of middleweight patients who continued to course 2
had an ANC of greater than or equal to 500 × 109/L,
which was not different from the 87.2% of overweight patients (P = .11) or 77.6% (P = .84)
of underweight patients. At 7 weeks, the respective proportions were 97.2%,
98.9%, and 100%. The time to recovery of ANC of 1000 × 109/L after the start of course 1 for the 3 groups was plotted (Figure 3): overweight and middleweight patients
had significantly faster neutrophil recovery than underweight patients (P = .004).
The time to complete a course of therapy was another indicator of toxicity. Figure 3 shows duration of course 1 in the 3
groups: there was no significant difference in the duration of the course.
When patients who had dose reductions were excluded, the relative positions
of the 3 groups were unchanged.
This study shows that overweight and underweight children and adolescents
with AML are less likely to survive than patients with BMI in the 11th through
94th percentiles. Inferior survival in both extreme BMI groups is attributable
to early treatment-related mortality, and treatment-related mortality is mostly
from infection. Although there is already substantial evidence that underweight
children with acute lymphoblastic leukemia and solid tumors experience increased
relapses and reduced survival,16- 19 this
is the first study to our knowledge to show excess mortality in overweight
pediatric cancer patients.
These results contrast with those in most adult cancers in which underweight
patients have no excess mortality and overweight patients have excess cancer-related
death rather than death from excessive toxicity.20- 25 One
notable exception is marrow transplantation: 3 studies show excess mortality
in obese adults from a combination of relapse and treatment-related mortality26,27 or treatment-related mortality alone.28 Dickson et al28 also found that underweight
patients experienced higher treatment-related mortality. Marrow transplantation
and contemporary AML therapy have in common dose-intensive chemotherapy complicated
by a relatively high baseline treatment-related mortality that is exaggerated
among underweight and overweight patients.
Malnutrition is associated with advanced disease, lower socioeconomic
status, immunodeficiency, increased number and spectrum of infections, reduced
access to care, and delays in diagnosis.18,19,29 Even
after adjustment for socioeconomic status, malnutrition continues to be associated
with poor outcome. Malnutrition in children reduces absorption, decreases
drug-protein binding, and impedes oxidative and other metabolic reactions.
These effects increase half-life, reduce clearance, and impair glomerular
filtration of drugs.30 They also augment toxicity.30 There is no information about the pharmacology of
cancer chemotherapy in underweight patients. Busulfan is the only drug for
which there is dosing information according to weight or BMI distribution
in children.31 Malnutrition reduces survival
in children with cancer in direct proportion to the extent of their malnutrition.18,19,29 Correction of nutritional
status improves outcomes.29 Thus, it would
seem reasonable to determine whether delaying therapy to initiate nutritional
supplementation and correction of immunodeficiency is possible, and if correction
is possible, to determine whether it improves outcome.
There is no obvious solution to the problem of excess treatment-related
mortality in overweight patients. The data concerning neutrophil recovery
and duration of course do not support the hypothesis that overweight patients
in this study received too much chemotherapy. There are few traditional pharmacologic
studies of bioavailability of chemotherapeutic agents in obese adults and
none in overweight children or adolescents. Several small studies have investigated
bioavailability of cyclophosphamide, ifosfamide, doxorubicin, and its metabolite
doxorubicinol as single agents.32- 34 Combination
chemotherapy has been investigated in 1 obese patient: compared with normal-weight
patients, she showed a substantially increased concentration over time for
4-hydroxy cyclophosphamide, thiotepa, and carboplatinum.35 Because
in general these pharmacologic studies show a trend for reduced clearance
and longer half-lives of chemotherapeutic agents, they seem to contradict
studies of antibiotics in obese adults in which hyperfiltration increases
clearance.36 Because most recent studies show
reduced rather than excessive toxicity among obese individuals, today the
consensus is that obese adult cancer patients receive too little rather than
too much chemotherapy.21- 25
In this study, overweight patients experienced more severe abdominal
pain, hypertension, pulmonary dysfunction, and coagulopathy. Unfortunately,
the information collected does not indicate when these toxicities occured,
how long they lasted, or whether they contributed to death or reflected end-stage
deterioration. Comorbidities such as these increase risk of death in adult
cancer patients with febrile neutropenia, so they may be important in these
patients.37 Overweight patients can also manifest
subtle immunologic abnormalities that could contribute to excess infectious
death.38,39 In 1970, Wiernik and
Serpick40 described 8 morbidly obese patients among 106 adults
with AML. None of the 8 patients survived longer than 1.75 months compared
with a median survival of 3.5 months for the whole study. These authors postulate
that subclinical diabetes, difficulties in performing thorough physical examinations,
or nuances in carrying out routine nursing care could contribute to early
This study has the limitations of a retrospective study: CCG-2961 was
not designed to investigate BMI as a variable. Some findings, such as the
proportion of overweight and underweight patients with donors for marrow transplantation,
may be spurious. The study does not illuminate the causes of excess infectious
treatment-related mortality. There are no socioeconomic data. All assessments
pertain to weight at diagnosis; it was not possible to determine whether weight
gain in underweight patients or weight loss in overweight patients improved
outcomes after the first course. Finally, it is likely that the observation
of excessive treatment-related mortality in overweight patients will be reproducible
only in pediatric cancers that involve extremely dose-intensive chemotherapy
or marrow transplantation studies as in obese adults.
The effect of BMI on outcome in pediatric AML is not a trivial problem:
the reduced survival in underweight and overweight patients is roughly equal
to the improved survival accomplished by 10 years of progress in pediatric
AML. Treatment-related mortality is the worst possible outcome for an individual
enrolled in a clinical trial, and if treatment-related mortality is not countered
by a net gain in survival, excess treatment-related mortality is also the
worst possible outcome for a clinical trial. These results have implications
for clinicians and clinical investigators. This is the first example of immediate
rather than impending life-shortening effects of excess weight in the young
and a confirmation of the risks of undernutrition in other pediatric cancers.16- 19 Interventions
currently available that could reduce the treatment-related mortality in underweight
and overweight groups include formal nutritional and immunologic assessment
at diagnosis. Underweight patients could benefit from preemptive nutritional
intervention or intravenous γ-globulin. In overweight patients, correction
of persistent moderate hyperglycemia and hypertension may remediate 2 important
comorbidities. Maintaining blood glucose concentration between 80 and 120
mg/dL appears to reduce mortality in patients in intensive care units.41
However, systematic changes in management should take place as part
of controlled studies. Basic pharmacokinetic and pharmacodynamic studies of
chemotherapeutic drug disposition in underweight and overweight patients are
likely to provide a rational basis for dosing. Until such information is available,
it is impossible to know whether doses of chemotherapy should be reduced.
Dose reduction is likely to lead to increased relapse, but relapse is the
lesser of 2 evils. It is also possible that dose reduction increases relapse
and has no effect on toxic mortality. Finally, technologists, nurses, and
physicians should examine whether overweight patients are receiving suboptimal
care because of difficulties in assessing them, as suggested by Wiernik and
Serpick.40 If that is the case, then it must
be determined as to how to change practice to overcome these barriers.
Corresponding Author: Beverly J. Lange,
MD, Division of Oncology, Children's Hospital of Philadelphia, 34th and
Civic Center Boulevard, Philadelphia, PA 19104 (firstname.lastname@example.org).
Author Contributions: Drs Lange and Alonzo
had full access to all of the data in the study and take responsibility for
the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lange, Feusner, Skolnik,
Sacks, Smith, Alonzo.
Acquisition of data: Lange, Feusner, Sacks,
Analysis and interpretation of data: Lange,
Gerbing, Feusner, Skolnik, Smith, Alonzo.
Drafting of the manuscript: Lange, Gerbing,
Skolnik, Sacks, Smith, Alonzo.
Critical revision of the manuscript for important
intellectual content: Lange, Feusner, Skolnik, Sacks, Smith, Alonzo.
Statistical analysis: Lange, Gerbing, Skolnik,
Obtained funding: Lange.
Administrative, technical, or material support:
Lange, Feusner, Skolnik, Sacks, Smith.
Study supervision: Lange, Feusner, Smith.
Funding/Support: This study was supported by
grants CA098543 and CA098413 from the National Institutes of Health (NIH)
from 2000 to 2004 and NIH U-10 grants listed at
from 1996 to 2000. Dr Lange was supported by the Yetta Dietch Novotny
Chair in Clinical Oncology. Chiron provided interleukin 2 to the National
Cancer Institute (NCI), which provided interleukin 2 to the institutions.
Role of the Sponsors: The Children’s
Cancer Group (CCG) was funded by an NIH U-10 grant to perform clinical trials
and correlative biology studies and epidemiologic studies in childhood cancer.
CCG-2961 was designed by the CCG-2961 Study Committee, which Dr Lange chaired.
The Committee reported to the CCG-AML Strategy Group. The design and conduct
of the study were reviewed by a data and safety monitoring board composed
of CCG and non-CCG cooperative group investigators, an ethicist, a lay member,
and representatives of the Clinical Trials Evaluation Program of the NCI,
who were nonvoting members. Primary patient data were obtained by nurses and
physicians; most were CCG members who were not compensated by CCG for their
work. Clinical research associates, supported fully or in part by the CCG
institutional grant, abstracted the data and transmitted them to the CCG Operations
office, Arcadia, Calif. Dr Lange received copies of all data capture forms.
Primary data analyses were performed by Dr Alonzo and Mr Gerbing, CCG statisticians.
The interpretation of the data was the primary responsibility of these 2 statisticians
and Dr Lange, with all authors participating substantially in the interpretation
of the data. All authors except Dr Skolnik, Ms Sacks, and Mr Gerbing were
members of the CCG-2961 committee. In 2000, CCG and 3 other groups conducting
clinical research in pediatric oncology fused to become the Children’s
Oncology Group (COG). The manuscript was sent to COG for final review by the
COG editor, Shaun Mason. Submission of the manuscript by Mr Mason to the journal
constitutes approval of the manuscript by COG. Chiron had no role in the design,
conduct, or analysis of the study.
Acknowledgment: We thank Christine Curran for
typing the manuscript and Shaun Mason, BA, for editorial support.