Content uploaded by Yuri Quintana
Author content
All content in this area was uploaded by Yuri Quintana on Mar 03, 2020
Content may be subject to copyright.
Childhood Cancer Epidemiology in
Low-Income Countries
Scott C. Howard, MD, MSc
1,2
Monika L. Metzger, MD, MSc
1,2
Judith A. Wilimas, MD
1,2
Yuri Quintana, PhD
2
Ching-Hon Pui, MD
1,2
Leslie L. Robison, PhD
3
Raul C. Ribeiro, MD
1,2
1
Department of Oncology, St. Jude Children’s
Research Hospital; Department of Pediatrics, Uni-
versity of Tennessee College of Medicine, Mem-
phis, Tennessee.
2
International Outreach Program, St. Jude Chil-
dren’s Research Hospital, Memphis, Tennessee.
3
Department of Epidemiology and Cancer Con-
trol, St. Jude Children’s Research Hospital; De-
partment of Pediatrics, University of Tennessee
College of Medicine, Memphis, Tennessee.
Global studies of childhood cancer provide clues to cancer etiology, facilitate pre-
vention and early diagnosis, identify biologic differences, improve survival rates
in low-income countries (LIC) by facilitating quality improvement initiatives, and
improve outcomes in high-income countries (HIC) through studies of tumor biol-
ogy and collaborative clinical trials. Incidence rates of cancer differ between vari-
ous ethnic groups within a single country and between various countries with
similar ethnic compositions. Such differences may be the result of genetic predis-
position, early or delayed exposure to infectious diseases, and other environmen-
tal factors. The reported incidence of childhood leukemia is lower in LIC than in
more prosperous countries. Registration of childhood leukemia requires recogni-
tion of symptoms, rapid access to primary and tertiary medical care (a pediatric
cancer unit), a correct diagnosis, and a data management infrastructure. In LIC,
where these services are lacking, some children with leukemia may die before diag-
nosis and registration. In this environment, epidemiologic studies would seem to
be an unaffordable luxury, but in reality represent a key element for progress. Hos-
pital-based registries are both feasible and essential in LIC, and can be developed
using available training programs for data managers and the free online Pediatric
Oncology Networked Data Base (www.POND4kids.org), which allows collection,
analysis, and sharing of data. Cancer 2008;112:461–72. 2007 American Cancer
Society.
KEYWORDS: high income countries, middle income countries, low income countries,
cancer epidemiology, developing countries, tumor registries.
The 5-year event-free survival for children with cancer is 75% to
79% in high-income countries (HIC).
1–3
However, 80% of the
world’s children live in middle- and low-income countries (MIC and
LIC), where poverty, lack of public health infrastructure, high mor-
tality rates in children under the age of 5 years (under 5-year mor-
tality rates), and low childhood cancer cure rates are pervasive. In
such settings, studies of cancer epidemiology may seem to be an
unaffordable luxury, but analysis of the global epidemiology of
childhood cancer and differences between LIC, MIC, and HIC is not
merely an academic exercise. Studies of childhood cancer in differ-
ent regions provide clues to cancer etiology, facilitate improvements
in public health through prevention and early diagnosis, identify
biologic differences that may require different therapeutic strategies,
improve survival rates in LIC by identification of causes of treatment
failure so that quality improvement initiatives can focus on these
causes, and improve outcomes in HIC through studies of tumor
biology and collaborative clinical trials (Table 1). Geographic differ-
ences in incidence may suggest unique genetic or environmental
exposures that affect cancer risk. In this report we review the inci-
Address for reprints: Scott C. Howard, MD,
St. Jude Children’s Research Hospital, 332 N.
Lauderdale St., Memphis, TN 38105-2794; Fax:
(901) 495-2099; E-mail: scott.howard@stjude.org
Supported in part by a Cancer Center Support
Grant (CA21765) from the National Cancer Insti-
tute and by the American Lebanese Syrian Asso-
ciated Charities (ALSAC). Dr. Pui is an American
Cancer Society Professor.
Received June 6, 2007; revision received August
15, 2007; accepted August 17, 2007.
ª2007 American Cancer Society
DOI 10.1002/cncr.23205
Published online 10 December 2007 in Wiley InterScience (www.interscience.wiley.com).
461
TABLE 1
Importance of Childhood Cancer Epidemiology in Low-Income Countries
Use of epidemiologic data Description Examples and references
Public health Programs for prevention and screening Education efforts can be targeted at the populations at highest risk; programs can raise awareness
at a societal level so that families and healthcare professionals work together to implement
screening and promote early diagnosis
Retinoblastoma in Honduras
27
Health planning Measurement of the geographic distribution and total number of cases of each cancer type allows
planning of the location where pediatric cancer units and satellite clinics should be established
and determination of services needed at each site
Honduras satellite clinics; development of
a regional flow cytometry center
45
Quality improvement Measuring outcomes of treatment and cancer-specific mortality identifies services that need to be
improved and facilitates assessment of the efficacy of interventions
Development of a pediatric cancer center
of excellence in Recife, Brazil
34
Clinical research Adaptation of pediatric oncology treatment
regimens
Conducting clinical trials of therapy in LIC that use less toxic, less expensive, or otherwise
modified versions of published treatment regimens can evaluate the feasibility and outcomes in
the local setting
ALL and lymphoma in Recife, Brazil,
33,34
Indonesia,
35
and India
46–48
Clinical research that can only be
performed in LIC
Evaluation of clinical problems unique to children with cancer in LIC, including abandonment of
therapy, the effects of extreme poverty on compliance and toxicity, and the effects of comorbid
illnesses (e.g. malnutrition, parasitic infection) on outcomes can lead to specific mitigation
strategies
Abandonment risk factors for children with
ALL
23,34,35
; telemedicine in Jordan to
improve treatment of central nervous
system cancer
49
Comparative clinical research Evaluation of specific aspects of care in diverse settings, such as the effects on outcome of culture,
language, socioeconomic status, and other variables that differ greatly between countries
Perceptions of pain in children with cancer
in Jordan
50
Clinical trials for patients with advanced
disease at diagnosis
Evaluation of treatment regimens in patients with high-stage disease at diagnosis to determine the
optimal treatment strategy in the local setting, where intense chemotherapy and stem cell
transplantation may not be feasible. Trials of new agents are also appropriate in such settings,
where the event-free survival without novel therapy is close to 0%. Indeed, the patients that
stand to gain most from novel therapies are those in LIC, where late diagnosis increases the
proportion of patients with incurable cancer
Extraocular retinoblastoma
27
Collaborative trials with global
participation
Multi-center, multi-national research on rare tumors with participation of centers in HIC, MIC and
LIC allow sufficient sample size to perform randomized controlled trials of therapy. Global
collaboration permits more rapid progress in therapeutics than if clinical trials are performed
only in HIC, where only 20% of children with cancer live
International study of infant ALL
(Interfant), haploidentical stem cell
transplantation in Chile
51
Epidemiology research Cancer etiology Assessment of differences in genetics, lifestyle, and environmental exposures between LIC and MIC
that correlate with different cancer incidence
Adrenocortical carcinoma in southern
Brazil
11
Cancer diagnosis Assessment of relative incidences of each type of childhood cancer to determine whether these
reflect genetic or environmental differences, or bias based on the differential probability of
survival until diagnosis among different cancers
ALL in Honduras
23
Basic research Discovery of new causes of childhood
cancer
Observation of unusual patterns of disease presentation or clusters of cancer within families or
regions may elucidate novel genetic and environmental risk factors for childhood cancer
Adrenocortical carcinoma in southern
Brazil
11,24,25
Biology research with primary tumor
samples
Patients with advanced disease have large tumors sufficient for a variety of biologic studies. In
HIC, less than 5% of retinoblastoma is extraocular, and the majority of tumors are not biopsied.
In contrast, 43% to 73% of retinoblastomas in LIC are extraocular and biopsy material is
available
27,52
Extraocular retinoblastoma
27,52
;
Adrenocortical carcinoma in Southeast
Brazil
11,24,25
Comparative cancer biology research Comparison of clinical and biologic features of the same cancer in distinct regions may help
identify unique clinical features, causes, and possibly therapies
Burkitt lymphoma in North America, Latin
America, and Africa
9
HIC indicates high-income countries; MIC, middle-income countries; LIC, low-income countries; BFM, Berlin-Frankfurt-Munster cooperative study group; ALL, acute lymphoblastic leukemia; NHL, non-Hodgkin lymphoma.
462 CANCER February 1, 2008 / Volume 112 / Number 3
dence rates of childhood cancer in MIC and LIC, dis-
cuss possible reasons for different reported incidence
rates, provide examples of the importance of epide-
miologic studies in LIC and their practical impor-
tance to patients and society, and propose the
universal implementation of hospital-based cancer
registries in pediatric cancer units as a feasible next
step to improve childhood cancer care and global
epidemiologic research.
Sources and Quality of Childhood Cancer
Epidemiology Data
Information about childhood cancer incidence in LIC
comes from hospital-based registries, population-
based registries, international organizations, and spe-
cific research projects. The International Agency for
Cancer Research (IARC) has conducted extensive stu-
dies of childhood cancer incidence throughout the
world by combining information from multiple popu-
lation-based tumor registries, and its publications
provide a comprehensive source of information about
cancer epidemiology in selected LIC (Table 2).
4,5
Causes of variation in cancer incidence rates
Differences in cancer incidence rates between HIC
and many LIC have been documented for childhood
cancer as a whole,
5–7
and for a variety of specific
cancers, including Burkitt and Hodgkin lympho-
mas,
8–10
adrenocortical carcinoma,
11,12
and acute
lymphoblastic leukemia (ALL), the most common
childhood cancer worldwide. International variation
in the incidence of ALL is well recognized.
13
Obser-
vations of a markedly increased incidence rate of
ALL in children between 2 and 5 years old in affluent
societies, the lack of such an age peak age in LIC,
and occasional clustering of childhood ALL cases
(especially in new towns) have fueled 2 parallel
infection-based theories of leukemogenesis: the
delayed-infection hypothesis
14
and the population-
mixing hypothesis.
15
Both hypotheses attribute the
peak incidence in industrialized countries to early in-
fectious insulation that predisposes the immune sys-
tem of susceptible individuals to aberrant or
pathologic responses after subsequent or delayed ex-
posure to common infections at an age commensu-
rate with increased lymphoid cell proliferation.
14,15
Some other cases of childhood ALL can be attributed
to maternal exposures during pregnancy,
16,17
in
which risk may be modulated by genetic polymorph-
isms of enzyme systems responsible for the metabo-
lism of drugs or environmental xenobiotics.
18–21
However, variations in environmental exposures and
genetic susceptibility can only account for small dif-
ferences in childhood leukemia incidence rates, and
do not explain the large differences (up to 10-fold)
between HIC and some LIC ( Table 2). Hence, the
role of underdiagnosis and underreporting must be
investigated.
Sources of error in estimating childhood cancer
incidence in LIC
Determination of cancer incidence requires both an
accurate estimate of the population of interest (eg,
younger than 15 years old) and an accurate count of
cancer cases within the population. Population esti-
mates depend on the accuracy and frequency of cen-
suses. Age-specific population estimates between
censuses are calculated by interpolation. However,
this approach does not provide valid estimates if the
accuracy of the most recent census is poor or when
there are large shifts in the population due to migra-
tion, refugees, or rapid changes in birth or death rates.
In such instances the age-specific population may be
over- or underestimated, and even determination of
the most likely direction of error may not be possible.
Compounding the problem of inaccurate population
estimates are potential errors in ascertainment and
characterization of cancer cases within the population
of interest. Cancer cases can only be considered if a
diagnosis is made and the case registered—a chain of
care that comprises several links (Fig. 1). In LIC, bar-
riers occur at all steps. Patients and parents may not
be aware of signs and symptoms of childhood cancer,
may rely on nonmedical forms of treatment, and may
not have the transportation or money to travel to a
primary care facility. If the patient arrives to primary
care, personnel may not be trained to recognize child-
hood cancer, laboratory and diagnostic imaging
equipment may not be available to screen for cancer,
and the patient or clinic may lack money to pay for
necessary testing and treatment. Similar barriers
make access to tertiary care and correct diagnosis
problematic, and even when correct diagnoses of can-
cer are made they may not be documented systemati-
cally in a cancer registry.
Any missing link in the chain of cancer diagnosis
can prevent ascertainment of the case, and cause the
reported cancer incidence rate to be lower than the
actual incidence rate—assuming an accurate popula-
tion census. The degree of underestimation depends
on many social, economic, and medical factors. In
countries like Jordan, with a total population of
5,900,000, a few major hospitals treat almost all chil-
dren in the country who develop cancer. It can be
argued that combining hospital registries of these
cancer centers approximates a population-based reg-
istry; however, even in Jordan there is a higher meas-
ured incidence rate in the capital city, which
Childhood Cancer Epidemiology in LIC/Howard et al. 463
TABLE 2
Incidence of Childhood Cancer per Million Children Less Than 15 Years Old in Selected Countries Categorized by Mean
per Capita Gross National Income
Country
Cancer
incidence
Leukemia
incidence
Nonleukemia
incidence
Gross National
income*
Total healthcare
spending*
Under 5-y
mortality rates
Low-income countries (n 59) 102 16 85 491 21 128
Malawi 100.0 1.1 98.9 160 13 175
Uganda 183.5 10.3 173.2
y
280 18 138
Zimbabwe 111.2 22.8 88.4
y
340 40 129
Mali 77.4 4.0 73.4 380 9 219
Nigeria 71.2 8.6 62.6 560 22 197
Vietnam 108.4 33.4 75.0 620 26 19
Papua New Guinea 100.0 8.1 91.9 660 23 93
Pakistan 100.0 40.5 59.5 690 13 101
India 64.4 19.2 45.2 730 27 85
Middle-income countries (n 518) 107 37 70 4537 241 25
Lower middle-income countries (n 58) 93 37 56 2324 93 33
Philippines 100.4 47.9 52.5 1300 31 33
China 104.8 40.2 64.6 1740 61 31
Ecuador 124.4 55.4 69.0 2180 109 26
Colombia 121.8 41.7 80.1 2290 138 21
Peru 104.4 35.6 68.8 2610 98 29
Algeria 69.6 37.3 32.3 2730 89 40
Thailand 70.1 28.1 42.0 2750 76 21
Namibia 45.6 6.2 39.4 2990 145 63
Upper middle-income countries (n 510) 118 37 81 6307 358 18
Bulgaria 98.6 32.0 66.6 3450 191 15
Brazil 100.0 27.8 72.2 3460 212 34
Uruguay 117.4 43.2 74.2 4360 323 14
Costa Rica 134.0 56.5 77.5 4590 305 13
South Africa 100.0 22.0 78.0 4960 295 67
Poland 111.0 35.0 76.0 7110 354 8
Slovakia 125.6 35.0 90.6 7950 360 9
Croatia 162.6 41.5 121.1 8060 494 7
Estonia 123.5 35.6 87.9 9100 366 8
Hungary 103.4 36.5 66.9 10030 684 6
High-income countries (n525) 130 41 89 32872 2516 5
Korea 106.4 36.9 69.5 15830 705 6
Portugal 146.7 36.0 110.7 16170 1348 5
Slovenia 113.5 36.3 77.2 17350 1218 4
Israel 131.0 25.2 105.8 18620 1514 6
United Arab Emirates 100.0 43.7 56.3 23770 661 8
Kuwait 109.7 32.3 77.4 24040 580 12
Spain 132.3 40.8 91.5 25360 1541 5
New Zealand 147.6 39.5 108.1 25960 1618 6
Germany 125.9 34.2 91.7 26220 3204 5
Singapore 125.3 48.2 77.1 27490 964 3
Hong Kong 128.9 52.4 76.5 27670 . 3
Italy 134.1 44.3 89.8 30010 2139 5
Australia 137.0 46.7 90.3 32220 2519 6
Canada 144.2 48.1 96.1 32600 2669 6
France 129.8 38.2 91.6 34810 2981 5
Netherlands 132.8 38.6 94.2 36620 3088 5
Finland 148.6 47.3 101.3 37460 2307 4
United Kingdom 118.2 38.6 79.6 37600 2428 6
Japan 107.6 35.5 72.1 38980 2662 4
Sweden 149.4 45.6 103.8 41060 3149 4
USA 137.9 43.1 94.8 43740 5711 8
Iceland 109.0 37.2 71.8 46320 3821 3
(continued)
464 CANCER February 1, 2008 / Volume 112 / Number 3
suggests that children with cancer in rural or distant
areas may have less access to diagnosis and treat-
ment.
22
We observed a similar pattern in Honduras
(population 7,500,000), where the measured annual
incidence of ALL in the capital city was 20 per mil-
lion versus 10 per million in distant and rural pro-
vinces.
23
These problems are probably even more
significant in larger countries, where changes in
referral patterns may not respect boundaries estab-
lished for the population used as the denominator
for incidence calculations.
Reported Versus Actual Incidence Rates
of Childhood Cancers
The difference in reported versus actual incidence
rates of childhood cancer is most extreme for leuke-
mia, a disease with protean signs and symptoms that
resemble those of infection, in which early death can
occur before cancer is suspected or diagnosed. By
contrast, lymphomas and solid tumors typically pres-
ent with a visible mass or other manifestation that
prompts parents to seek medical care. Furthermore,
early death due to lymphomas and solid tumors is
less common, even when the disease reaches an
advanced stage. The mean annual leukemia incidence
per million children was 16.4 (standard deviation [SD]
13.6) in LIC, 36.5 (SD 11.6) in MIC, and 40.9 (SD 6.1)
in HIC (Table 2), an observation that supports the
contention that leukemia incidence is systematically
underestimated in LIC (Fig. 2). In contrast, the inci-
dence of nonleukemia cancers was 85 (SD 37) in LIC,
70 (SD 20.5) in MIC, and 89 (SD 14) in HIC (Table 2),
which does not support a pattern of systematic
underestimation of nonleukemias in LIC (Fig. 3). After
exclusion of Kaposi sarcoma, which is common in
Uganda and Zimbabwe, the incidence rates of non-
leukemia cancers in LIC decreases to 76. LIC with the
lowest reported incidence rates of leukemia have a
very high incidence of malaria (>200 cases per 1000
population per year), suggesting that patients with
leukemia may die with anemia and fever that is attrib-
uted to malaria, which is 10,000 times more common
than leukemia in endemic areas.
Competing causes of death in LIC
One proposed cause of lower reported incidence of
childhood cancer in LIC is the high mortality rate
among children younger than 5 years of age in some
countries, which may lead to death of a child before
development of cancer. However, premature death of
FIGURE 1. Links in the chain of childhood cancer diagnosis and registration. Many steps are required for a child with cancer to be diagnosed and registered.
In low-income countries, barriers occur at all steps. SES indicates socioeconomic status.
TABLE 2
(continued)
Country
Cancer
incidence
Leukemia
incidence
Nonleukemia
incidence
Gross National
income*
Total healthcare
spending*
Under 5-y
mortality rates
Denmark 149.3 47.2 102.1 47390 3534 5
Switzerland 139.5 43.8 95.7 54930 5035 5
Norway 143.2 44.0 99.2 59590 4976 4
Incidence data are from the International Agency for Research on Cancer.
5
Low-income country (LIC) is defined as a country in which the mean per capita annual income in 2005 is less than US $825; middle-
income country (MIC) is a country in which the mean per capita annual income is $825 to $10,065. MIC are divided into lower middle-income country (mean per capita annual income of $825 to $3255) and
upper middle-income country (mean per capita annual income of $3256 to $10,065); high-income country (HIC) is a country in which the mean per capita annual income is more than $10,065.
* Annual per capita figures in US dollars. Gross national incomes were taken from the world development indicators database of the World Bank for 2005.
y
Kaposi sarcoma accounted for 68.5 nonleukemia cancers per million per year in Uganda and 10.7 in Zimbabwe.
Childhood Cancer Epidemiology in LIC/Howard et al. 465
FIGURE 2. Relation of the reported incidence rate of childhood leukemia to gross national income. The reported incidence of childhood leukemia (all types
combined) varies significantly according to mean annual per capita gross national income (GNI). In low-income countries there is a wide range of recorded leu-
kemia incidence. This range is much narrower in upper middle-income countries, which report an average of 37 cases per million children per year, and high-
income countries, which report an average of 41 cases per million per year. In low-income countries, the reported incidence rate of leukemia correlates with
GNI (r50.56, P5.12), but less so in middle- (r520.05, P5.83) and high-income countries (r50.38, P5.06).
FIGURE 3. Relation of the reported incidence rate of childhood nonleukemia cancers to gross national income (GNI). The reported incidence of nonleukemia
childhood cancers does not vary consistently according to the category of mean annual per capita GNI, although there is a weak positive correlation of GNI
with nonleukemia cancer incidence when all groups are combined (r50.31, P5.02). Uganda, which has an annual incidence of 173.2 nonleukemia childhood
cancers per million and a GNI of $280, is not shown.
466 CANCER February 1, 2008 / Volume 112 / Number 3
children due to infection and malnutrition does not
change the incidence rate very much, because such
deaths are presumed to occur in equal proportion in
children who would have later developed cancer as in
those who would not have done so. In other words, if
10% of children die before reaching the age of 5 years,
there will be 10% fewer cancer cases among children
aged 6 to 15 years, but there will also be 10% fewer
children without cancer in this age group, so the inci-
dence rate will remain unchanged. Of the 52 countries
reported in Table 2, Mali has the highest under 5-year
mortality: 219 per 1000 children (21.9%). Most of
these children die of infection, whose symptoms
resemble those of leukemia. For this reason, a high
under 5-year mortality correlates strongly with a
lower reported incidence of leukemia (Fig. 4, P<.001),
because regions in which young children die from
infection are the same as those in which children with
leukemia will die before diagnosis. In contrast, the
reported incidence rate of nonleukemia cancers
does not correlate with under 5-year mortality (Fig. 5,
P5.89), because children with solid tumors do not
die of infection before diagnosis.
Adrenocortical Carcinoma and Retinoblastoma as Models
of the Usefulness of International Childhood Cancer
Epidemiology
Adrenocortical carcinoma (ACC)
The estimated annual incidence of ACC in the US is
0.3 per million children younger than 15 years of
age.
5
The disease often occurs in association with
the Li-Fraumeni familial cancer syndrome, in which
mutations in the germline p53 gene predispose to a
variety of cancers, including ACC. However, in the
Parana and Sao Paulo states of southern Brazil the
incidence of ACC is 10–15 times greater, but no
endemic infections, environmental or occupational
exposures, ethnic predisposition, or kindreds with
Li-Fraumeni syndrome could be identified.
12
From
1996 to 1999 a subset of 92 children with ACC trea-
ted at a single institution in southern Brazil under-
went genotyping of p53 and were found to have an
identical point mutation in exon 10 encoding an ar-
ginine-to-histidine amino acid substitution at codon
337 of p53.
24
Although half of first-degree and a
third of second-degree relatives had a similar point
mutation, there was no family history of cancers to
suggest Li-Fraumeni syndrome.
11
Functional studies
of the protein derived from the mutated p53 gene
revealed that p53 in these patients had normal ac-
tivity except at high pH, which can be found in the
adrenal cortex in its physiologic state, a finding that
partially explained the tissue-specific cancer predis-
position.
25
A tumor registry for ACC has now been
established to facilitate continued clinical and bio-
logic studies and to prepare an infrastructure for
subsequent clinical trials of prevention and early
detection.
26
Retinoblastoma
Similarly, recognition of the apparent high incidence
and advanced stage of presentation of retinoblastoma
FIGURE 4. Relation of the reported incidence rate of childhood leukemia to under 5-year mortality in low- and middle-income countries. In low- and middle-
income countries the reported incidence of childhood leukemia (all types combined) rises as the under 5-year mortality decreases (r520.78, P<.001). This
inverse correlation reflects improved survival until diagnosis and registration as under 5 mortality decreases.
Childhood Cancer Epidemiology in LIC/Howard et al. 467
in Honduras led to development of programs to pro-
mote universal screening, early diagnosis, improve-
ments in treatment, collaborative studies of tumor
biology with scientists at St. Jude Children’s Research
Hospital, and a multinational clinical trial of therapy
in Central America.
27,28
In Honduras from 1995 to
2003, 73% of children presented with extraocular dis-
ease. A national retinoblastoma education campaign
was undertaken in concert with a national vaccination
effort, and the rate of extraocular disease decreased to
35% in the subsequent 2 years (P5.002).
27
Extended educational programs are under way to
further reduce diagnostic delays, but new treatments
are needed for children who present with extraocular
disease. To develop such treatments an improved
understanding of retinoblastoma biology is needed.
Studies in cell lines and mice have been very promis-
ing in this regard,
29–32
but clinical trials in humans
will be needed to definitively test any new drug or
combination. In the US an estimated 200 children
per year develop retinoblastoma, but only about 10
of these (5%) have extraocular disease at diagnosis.
Clinical trials of new agents will require large-scale
international cooperation, and the participation of
centers in LIC will be critical (Table 1). Such trials
should be performed in concert with community
education programs to promote early diagnosis, in
pediatric cancer centers with expertise in both chem-
otherapy administration and ocular local control
measures, such as those being developed in Panama,
Honduras, and Guatemala.
27,28
The Way Forward
Improving survival rates in LIC
The only way to know with certainty the optimal
treatment strategy in a particular LIC setting is to
implement uniform, protocol-based care for each
childhood cancer and to carefully monitor rates of
toxic death, abandonment of treatment, and recur-
rence.
28,33–35
In some cases adjustment of the chem-
otherapy regimen will be required to maximize the
probability of cure; in all cases, improvements in
supportive care and efforts to reduce abandonment
will be required.
35
A hospital-based cancer registry
and active data management program are essential
to successfully monitor outcomes and measure the
effect of specific interventions.
36
Such registries must
always be developed in the context of a pediatric
cancer unit, which serves as the focal point for
efforts to improve the quality of care.
Hospital cancer registries are feasible everywhere
In light of the difficulties and costs associated with
accurate population-based cancer registries and sig-
nificant uncertainties in estimates of both cancer
cases and the populations from which they derive,
how can LIC obtain accurate epidemiologic informa-
tion and take advantage of its many benefits to
patients and society? The logical first step is the
implementation of a pediatric cancer unit coupled
with a registry in all hospitals where children with
cancer are treated. Such registries can be maintained
at low cost and comprise a key component of a pedi-
FIGURE 5. Relation of the reported incidence rate of childhood nonleukemia cancers to under 5-year mortality in low- and middle-income countries. In low-
and middle-income countries the reported incidence of nonleukemia cancers does not correlate with under 5-year mortality (r520.02, P5.89).
468 CANCER February 1, 2008 / Volume 112 / Number 3
atric oncology data management program that
includes a data manager, database, and data analysis.
They also serve as a practical first step toward a possi-
ble population-based assessment of cancer incidence
rates. Finally, an excessively high or low incidence
rate for a particular cancer may serve as an indicator
of systematic misdiagnosis of certain types of cancer,
particularly among cancers that are difficult to distin-
guish from each other without expertise and infra-
structure for pathologic diagnosis. For example, if
Ewing sarcoma were frequently misdiagnosed as
rhabdomyosarcoma, the registry would reveal an
unrealistically low rate of Ewing sarcoma and an
excessively high incidence of rhabdomyosarcoma,
and would suggest a need for further investigation.
Data managers
Data managers in LIC can be hired and trained at
low cost, with close supervision by local physicians
and extensive use of internet communication via the
free educational website www.Cure4Kids.org.
36
In
Honduras such a program was successfully imple-
mented with a 2-day onsite training workshop fol-
lowed by regular online communication and local
supervision. In many LIC data manager salaries are
less than US $600 per month plus the cost of a com-
puter with internet access. All necessary software is
available at no cost from the International Outreach
Program of St. Jude Children’s Research Hospital.
Weekly data manager training sessions are held via
www.Cure4kids.org in both English and Spanish, and
data managers from any country are welcome to par-
ticipate at no cost.
POND Database
The Pediatric Oncology Networked Database (POND,
www.Pond4kids.org) has been in use since 2004 and
is currently in its second version.
28,36,37
This multilin-
gual, secure, online database was designed for data
management programs in LIC, and is provided at no
cost. In addition to standard tumor registry, cancer-
specific, and toxicity information, POND can store
nutrition, psychosocial, and socioeconomic informa-
tion, which can be used to assess a patient’s risk for
abandonment of treatment. Chemotherapeutic regi-
mens can be stored in POND with automatic genera-
tion of patient-specific treatment ‘roadmaps’ and
calculation of chemotherapeutic drug doses. Protocols
can be shared via a global library so that other sites
can use them. POND allows sharing of automatically
deidentified data with local and international colla-
borators, hospital administration, government agen-
cies, and nongovernmental agencies for healthcare
planning, outcomes assessment, quality improvement,
and research. However, control of the data always
remains with the site administrator. Sharing can be
turned on or off at will, according to the site’s needs,
and a complete export of the data in the universal
.xml format can be performed should the site decide
that another software tool better meets its needs. The
system currently supports English, Spanish, French,
Portuguese, and Chinese, and is used at 33 sites in 16
countries, with more than 11,000 patients registered.
POND is used as the database for the American Soci-
ety of Hematology’s international acute promyelocytic
leukemia protocol and potentially could serve the
needs of other international study groups.
Data analysis
Although a data manager and database are essential
prerequisites, data analysis is the ultimate goal. Rapid
adoption of POND occurred because physicians and
hospital administrators saw immediate benefits to
real-time data collection and periodic analysis of
results. Analysis addresses local problems, such as
abandonment of treatment, which is the most com-
mon cause of treatment failure in LIC.
23,38
Doctors in
Guatemala generate the list of patients to be seen on
a particular day and at the end of the day review all
visits to make sure that patients who missed appoint-
ments can be contacted and encouraged to resume
therapy the next day. Social workers store key socioe-
conomic information that helps determine eligibility
for support programs, such as subsidized transporta-
tion for clinic visits. Hospital administrators use the
data to assess personnel needs (nursing, laboratory,
social work, etc) and in some cases the government is
provided information from POND to determine fund-
ing needs for the pediatric oncology program.
Collaborators at 8 centers in 7 Central American
countries who use shared treatment protocols imple-
mented via POND can assess toxicity and event-free
survival of protocol patients in real time using the
sharing mechanism (which can be limited to a speci-
fic disease or protocol). Kaplan-Meier curves can be
generated automatically by POND and other statisti-
cal analyses will be added in version 3 because many
clinicians in LIC do not have access to statistical
analysis programs and also lack resources to contract
an epidemiologist or statistician to assist with analy-
sis. Training in the conduct and analysis of clinical
trials, and review of patients with difficulties are con-
ducted via email and regular online conferences via
www.Cure4kids.org. In this regard, statisticians and
clinical researchers from the Monza International
School of Pediatric Hematology/Oncology (MISPHO)
and the Pediatric Oncology Group of Ontario (POGO)
have been particularly helpful.
28,37,39–41
Childhood Cancer Epidemiology in LIC/Howard et al. 469
Funding for data management programs in LIC
Data management programs in LIC are inexpensive,
and many stakeholders benefit from the data col-
lected.
42
However, initiation and maintenance of a
successful program does require some funding. In
many cases, nonprofit foundations that support pedi-
atric cancer units in LIC have used donated money to
fund data management programs, just as they do to
provide essential medications and subsidized trans-
portation.
34
The Central American program was initi-
ally funded by a 3-year grant from POGO, which paid
for data manager training and salaries. As part of its
My Child Matters program, Sanofi-Aventis and the
International Union against Cancer (UICC) funded
projects in LIC that included data management com-
ponents. International research agencies are another
potential source of support. An additional, as-yet
untapped resource may be the pharmaceutical indus-
try. One could even imagine branded data manage-
ment programs, in which the supporting company is
specifically recognized. Perhaps the most important
source of ongoing support is that provided by nonpro-
fit foundations in-country, which provides an oppor-
tunity for individuals to help fellow citizens and
creates local capacity. This model has proved very
successful in Recife, Brazil, and elsewhere.
34
Objections
Although pediatric cancer unit-based registries, data
management programs, and global epidemiology stu-
dies are feasible at modest cost, should they be a pri-
ority in LIC? Indeed, should cancer care be
supported at all in a country like Mali, where 22% of
children die before reaching 5 years of age, 47% of
births occur with no prenatal visit, only 41% are vac-
cinated for measles in rural areas, and 43% have
growth stunting from malnutrition (www.who.int/
whosis/whostat2007)? Even if every child with cancer
were cured of the disease, the under 5-year mortality
would decrease by less than 1 per 1000, so in coun-
tries like Malawi, Nigeria, and Mali, where the under
5-year mortality is 175 to 219 per 1000, clean water,
food, vaccines, antibiotics, oral rehydration pro-
grams, and malaria treatment remain the highest
health priorities. However, in countries where an
attempt is made to treat children with cancer, the
diagnosis and outcome of these children should be
recorded and analyzed so that use of limited
resources can be optimized in the local setting. In
this regard, 33% of 42 children with Burkitt lym-
phoma in Malawi were cured with a 42-day treat-
ment regimen adapted to local conditions with a
cost of US $250 per patient.
43
These patients were
cured (defined as remission for at least 12 months,
after which recurrence is uncommon in Burkitt lym-
phoma) despite the finding that 52% presented mal-
nourished and 39% with an active parasitic infection
(usually malaria). Whether the funds used for cancer
care would have saved more lives had they been
spent on other health problems is an important
question, but it is encouraging that even in the poor-
est regions some children with cancer can be cured
with existing resources under local conditions. In the
poorest LIC we do not advocate diversion of govern-
ment health funds away from essential services such
as vaccine and malaria control programs to treat
childhood cancer; however, because treatment of
curable illnesses is a fundamental right of children,
44
we would propose that members of the healthcare
system in LIC seek support for pediatric cancer care
from every possible source, including international
agencies and foundations, and that results be docu-
mented via the hospital-based registry.
Conclusions
In summary, the study of childhood cancer epidemi-
ology in LIC may seem like a relatively low priority
considering competing public health and medical
demands. However, the large number of children
with cancer in LIC, the need for health planning,
clinical research to adapt treatment regimens to local
conditions, and the opportunity for epidemiologic
research make pediatric cancer unit (hospital)-based
registries potentially cost-effective. Well-designed
and maintained hospital registries can be established
with modest financial support, represent an integral
component of pediatric cancer care, and provide a
potential platform for expansion to a population-
based registry when feasible. As registries are estab-
lished, the reported childhood leukemia incidence in
LIC in all likelihood will increase, reflecting an
improved healthcare infrastructure and providing an
important marker of societal progress.
ACKNOWLEDGMENTS
We thank the Pediatric Oncology Group of Ontario
(POGO) for supporting a data management program
in Central America that served as the prototype for
many of the ideas presented here and Hemalatha
Kundurthi for invaluable assistance with data collection.
REFERENCES
1. Jemal A, Siegel R, Ward E, Murray T, Xu J, Thun MJ. Cancer
statistics, 2007. CA Cancer J Clin. 2007;57:43–66.
2. Gatta G, Capocaccia R, Stiller C, Kaatsch P, Berrino F, Ter-
enziani M. Childhood cancer survival trends in Europe: a
EUROCARE Working Group study. J Clin Oncol. 2005;23:
3742–3751.
470 CANCER February 1, 2008 / Volume 112 / Number 3
3. Pui CH, Evans WE. Treatment of acute lymphoblastic leu-
kemia. N Engl J Med. 2006;354:166–178.
4. Parkin DM, Stiller CA, Draper GJ, Bieber CA. The interna-
tional incidence of childhood cancer. Int J Cancer. 1988;42:
511–520.
5. Parkin DM, Kramarova E, Draper GJ, et al. International
Incidence of Childhood Cancer. Vol II. Lyon, France: Inter-
national Agency for Cancer Research; 1998;1–391.
6. Miller RW. Geographical and ethnic differences in the
occurrence of childhood cancer. IARC Sci Publ. 1988;3–7.
7. Miller RW. Epidemiologic evidence for genetic variability in
the frequency of cancer: ethnic differences. Basic Life Sci.
1988;43:65–70.
8. Magrath IT. Non-Hodgkin’s lymphomas: epidemiology and
treatment. Ann N Y Acad Sci. 1997;824:91–106.
9. Rao CR, Gutierrez MI, Bhatia K, et al. Association of Bur-
kitt’s lymphoma with the Epstein-Barr virus in 2 develop-
ing countries. Leuk Lymphoma. 2000;39:329–337.
10. Nerurkar AY, Vijayan P, Srinivas V, et al. Discrepancies in
Epstein-Barr virus association at presentation and relapse
of classical Hodgkin’s disease: impact on pathogenesis. Ann
Oncol. 2000;11:475–478.
11. Figueiredo BC, Sandrini R, Zambetti GP, et al. Penetrance
of adrenocortical tumours associated with the germline
TP53 R337H mutation. J Med Genet. 2006;43:91–96.
12. Sandrini R, Ribeiro RC, DeLacerda L. Childhood adrenocor-
tical tumors. J Clin Endocrinol Metab. 1997;82:2027–2031.
13. McNeil DE, Cote TR, Clegg L, Mauer A. SEER update of
incidence and trends in pediatric malignancies: acute lym-
phoblastic leukemia. Med Pediatr Oncol. 2002;39:554–557.
14. Greaves M. Infection, immune responses and the aetiology
of childhood leukaemia. Nat Rev Cancer. 2006;6:193–203.
15. Kinlen L. Infections and immune factors in cancer: the role
of epidemiology. Oncogene. 2004;23:6341–6348.
16. Spector LG, Davies SM, Robison LL, Hilden JM, Roesler M,
Ross JA. Birth characteristics, maternal reproductive his-
tory, and the risk of infant leukemia: a report from the
Children’s Oncology Group. Cancer Epidemiol Biomarkers
Prev. 2007;16:128–134.
17. Spector LG, Xie Y, Robison LL, et al. Maternal diet and
infant leukemia: the DNA topoisomerase II inhibitor hypo-
thesis: a report from the Children’s Oncology Group.
Cancer Epidemiol Biomarkers Prev. 2005;14:651–655.
18. Skibola CF, Smith MT, Hubbard A, et al. Polymorphisms in
the thymidylate synthase and serine hydroxymethyltrans-
ferase genes and risk of adult acute lymphocytic leukemia.
Blood. 2002;99:3786–3791.
19. Healy J, Belanger H, Beaulieu P, Lariviere M, Labuda D,
Sinnett D. Promoter SNPs in G1/S checkpoint regulators
and their impact on the susceptibility to childhood leuke-
mia. Blood. 2007;109:683–692.
20. Lanciotti M, Dufour C, Corral L, et al. Genetic polymorph-
ism of NAD(P)H:quinone oxidoreductase is associated with
an increased risk of infant acute lymphoblastic leukemia
without MLL gene rearrangements. Leukemia. 2005;19:214–
216.
21. Pui CH, Relling MV. Topoisomerase II inhibitor-related
acute myeloid leukaemia. Br J Haematol. 2000;109:13–23.
22. Al Sheyyab M, Bateiha A, Kayed SE, Hajjawi B. The inci-
dence of childhood cancer in Jordan: a population-based
study. Ann Saudi Med. 2003;23:260–263.
23. Metzger ML, Howard SC, Fu LC, et al. Outcome of child-
hood acute lymphoblastic leukaemia in resource-poor
countries. Lancet. 2003;362:706–708.
24. Ribeiro RC, Sandrini F, Figueiredo B, et al. An inherited p53
mutation that contributes in a tissue-specific manner to
pediatric adrenal cortical carcinoma. Proc Natl Acad Sci
USA. 2001;98:9330–9335.
25. DiGiammarino EL, Lee AS, Cadwell C, et al. A novel mech-
anism of tumorigenesis involving pH-dependent destabili-
zation of a mutant p53 tetramer. Nat Struct Biol. 2002;9:
12–16.
26. Michalkiewicz E, Sandrini R, Figueiredo B, et al. Clinical
and outcome characteristics of children with adrenocorti-
cal tumors: a report from the International Pediatric Adre-
nocortical Tumor Registry. J Clin Oncol. 2004;22:838–845.
27. Leander C, Fu LC, Pena A, et al. Impact of an education
program on late diagnosis of retinoblastoma in Honduras.
Pediatr Blood Cancer. 2007;49:817–819.
28. Howard SC, Ortiz R, Baez LF, et al. Protocol-based treat-
ment for children with cancer in low income countries in
Latin America: a report on the recent meetings of the
Monza International School of Pediatric Hematology/On-
cology (MISPHO)–part II. Pediatr Blood Cancer. 2007;48:
486–490.
29. Dyer MA, Rodriguez-Galindo C, Wilson MW. Use of precli-
nical models to improve treatment of retinoblastoma. PLoS
Med. 2005;2:e332.
30. Laurie NA, Donovan SL, Shih CS, et al. Inactivation of the
p53 pathway in retinoblastoma. Nature. 2006;444:61–66.
31. Laurie NA, Gray JK, Zhang J, et al. Topotecan combination
chemotherapy in 2 new rodent models of retinoblastoma.
Clin Cancer Res. 2005;11:7569–7578.
32. Zhang J, Schweers B, Dyer MA. The first knockout mouse
model of retinoblastoma. Cell Cycle. 2004;3:952–959.
33. Hsu SC, Metzger ML, Hudson MM, et al. Comparison of
treatment outcomes of childhood Hodgkin lymphoma in 2
US centers and a center in Recife, Brazil. Pediatr Blood
Cancer. 2007;49:139–144.
34. Howard SC, Pedrosa M, Lins M, et al. Establishment of a
pediatric oncology program and outcomes of childhood
acute lymphoblastic leukemia in a resource-poor area.
JAMA. 2004;291:2471–2475.
35. Mostert S, Sitaresmi MN, Gundy CM, Sutaryo, Veerman AJ.
Influence of socioeconomic status on childhood acute lym-
phoblastic leukemia treatment in Indonesia. Pediatrics.
2006;118:e1600–e1606.
36. Ayoub L, Fu L, Pena A, et al. Implementation of a data
management program in a pediatric cancer unit in a low
income country. Pediatr Blood Cancer. 2007;49:23–27.
37. Howard SC, Marinoni M, Castillo L, et al. Improving out-
comes for children with cancer in low-income countries in
Latin America: a report on the recent meetings of the Monza
International School of Pediatric Hematology/Oncology
(MISPHO)-pt I. Pediatr Blood Cancer. 2007;48:364–369.
38. Mostert S, Sitaresmi MN, Gundy CM, Sutaryo , Veerman
AJ. Influence of socioeconomic status on childhood acute
lymphoblastic leukemia treatment in Indonesia. Pediatrics.
2006;118:e1600–e1606.
39. Barr RD, Gonzalez A, Longchong M, et al. Health status
and health-related quality of life in survivors of cancer in
childhood in Latin America: a MISPHO feasibility study. Int
J Oncol. 2001;19:413–421.
40. Barr RD, Ribeiro RC, Agarwal BR, Masera G, Hesseling PB,
Magrath I. Pediatric oncology in countries with limited
resources. In: Pizzo PA, Poplack DG, eds. Principles and
Practice of Pediatric Oncology. Philadelphia: Lippincott,
Williams and Wilkins;2006:1604–1616.
Childhood Cancer Epidemiology in LIC/Howard et al. 471
41. Sala A, Barr RD, Masera G. A survey of resources and activ-
ities in the MISPHO family of institutions in Latin America:
a comparison of 2 eras. Pediatr Blood Cancer. 2004;43:758–
764.
42. Naafs-Wilstra M, Barr R, Greenberg C, et al. Pediatric on-
cology in developing countries: development of an alliance
of stakeholders. Med Pediatr Oncol. 2001;36:305–309.
43. Hesseling P, Broadhead R, Mansvelt E, et al. The 2000 Bur-
kitt lymphoma trial in Malawi. Pediatr Blood Cancer. 2005;
44:245–250.
44. Pui CH, Schrappe M, Masera G, et al. Ponte di Legno
Working Group: statement on the right of children with
leukemia to have full access to essential treatment and
report on the Sixth International Childhood Acute Lympho-
blastic Leukemia Workshop. Leukemia. 2004;18:1043–1053.
45. Howard SC, Campana D, Coustan-Smith E, et al. Develop-
ment of a regional flow cytometry center for diagnosis of
childhood leukemia in Central America. Leukemia. 2005;19:
323–325.
46. Magrath I, Shanta V, Advani S, et al. Treatment of acute
lymphoblastic leukaemia in countries with limited
resources; lessons from use of a single protocol in India
over a twenty year period [Corrected]. Eur J Cancer.
2005;41:1570–1583.
47. Advani S, Pai S, Venzon D, et al. Acute lymphoblastic leu-
kemia in India: an analysis of prognostic factors using a
single treatment regimen. Ann Oncol. 1999;10:167–176.
48. Advani S, Pai S, Adde M, et al. Preliminary report of an
intensified, short duration chemotherapy protocol for the
treatment of pediatric non-Hodgkin’s lymphoma in India.
Ann Oncol. 1997;8:893–897.
49. Qaddoumi I, Mansour A, Musharbash A, et al. Impact of
telemedicine on pediatric neuro-oncology in a developing
country: the Jordanian-Canadian experience. Pediatr Blood
Cancer. 2007;48:39–43.
50. Forgeron PA, Finley GA, Arnaout M. Pediatric pain preva-
lence and parents’ attitudes at a cancer hospital in Jordan.
J Pain Symptom Manage. 2006;31:440–448.
51. Palma J, Mosso C, Paris C, et al. Establishment of a pediat-
ric HSCT program in a public hospital in Chile. Pediatr
Blood Cancer. 2006;46:803–810.
52. Bowman RJ, Mafwiri M, Luthert P, Luande J, Wood M. Out-
come of retinoblastoma in east Africa. Pediatr Blood Can-
cer. 2007 (in press). 2006 Nov 21; [Epub ahead of print]
472 CANCER February 1, 2008 / Volume 112 / Number 3
