Measures
The understanding of the
epidemiology of any medical problem demands the
use of basic terminology from the language of
statistics.
Ratio - Relationship between 2
quantities [x/y]
Proportion - Ratio in which the
denominator also includes the numerator [x/(y+x)]
Rate - Proportion occurring per
time period [x/(x+y)/time]
Incidence - Proportion of new
cases within a population over time [x/(y+x)/time]
Prevalence - Number of existing
cases in a population at a set time
Crude rate - Measure of actual
events in a population
Standardized rate - Crude rate
adjusted for a factor in the population (eg, age,
sex, economic status)
Standardized
mortality/incidence ratios - Observed rates
adjusted by comparison with the expected rate
derived from a large population
Relative risk - Incidence in a
population with a specific characteristic compared
to that in a population without the characteristic
Study designs
Descriptive design - Defines
the characteristics of a particular disease entity
Ecological design - Compares
large populations (eg, populations of nations)
Prospective design - Identifies
2 similar populations to be treated in different
ways in the future for subsequent analysis
Retrospective design -
Identifies and analyzes 2 similar populations that
were treated in different ways in the past
Clinical trials - Administered
chiefly under the auspices of national groups with
supervision by and interaction with the National
Cancer Institute (NCI) of the National Institutes
of Health (NIH) and the Food and Drug
Administration (FDA)
Clinical Drug Trials
New cancer drugs historically
have been adapted for pediatric use after use in
adult patients. Recently, new drug development has
incorporated pediatric trials that occur
concurrently with adult trials, which follow
research and development in private industry and
academia. Drug development is monitored by the
Cancer Therapy Evaluation Program (CTEP) of the
NIH.
A typical clinical trial
protocol includes the following information:
objectives of the trial, background, patient
eligibility criteria, study design, treatment
plan, drug information, treatment evaluation
criteria, data collection methods, plan for
statistical analysis, consent form to be signed by
the patient and investigator, and supporting
references. Relevant appendices also are attached.
Phase I trials
Phase I trials are designed
specifically to assess toxicity. Pediatric
patients are treated in cohorts of 3 starting at a
dose that is either 75% of the adult dose or 10%
of the lethal dose used in mouse studies. The dose
is increased in predetermined steps for each new
cohort of patients. Toxicity is assessed in
multiple body systems, and a level of
dose-limiting toxicity (DLT) is defined. If DLT
occurs in at least 1 of the 3 patients, the
protocol drops back to the previous dose level,
unless the DLT involves only hematologic toxicity
in a patient with a hematologic malignancy.
Phase II trials
Typically, phase II trials are
designed to assess the efficacy of a drug directly
in particular tumor types. A dose presumed to be
safe from the results of phase I trials is used.
An objective measure of response, such as
percentage decrease in tumor size by scan, is used
to evaluate efficacy. Typically, a 2-stage
process, which first attempts to establish firm
likelihood and then goes on to measure smaller
differences, is used.
Phase III trials
Phase III studies are intended
to test the efficacy of novel ways of using
accepted drugs (eg, combination chemotherapy,
neoadjuvant therapy, timing variations, dose
intensification) in comparison with standard
therapy or the natural history of the disease. The
design must measure and account for the potential
of false-positive and false-negative data.
Potential for error can be calculated and used to
decide on the number of patients who need to be
enrolled to ensure a certain level of confidence
in the results. A type I error occurs when the
P value, which is the probability of
obtaining the observed data (or data that are more
extreme) if the null hypotheses were exactly true,
suggests that a proposed treatment is better than
standard when it is not.
Phase III trial design can be
sequential, allowing data to be evaluated
continuously to find efficacious treatments as
quickly as possible. However, type I errors can
become magnified in this type of trial. This
phenomenon can be blunted by requiring greater
significance for the study. Factorial designs
examine multiple factors using a randomization
method. Equivalence trials can be designed to
demonstrate whether a treatment strategy of
reduced duration and dosage is equally as
efficacious as standard therapy.
The key to any phase III trial
is using randomization, which ensures that
patients are allocated to respective arms without
bias. A method of allocating patients based on
random numbers removes predictability from the
assignment. Stratification is also desirable in
order to group patients with identifiable
prognostic characteristics. Ideally, such
assignments are conducted at the onset of therapy.
A careful analysis of the
protocol and a clear understanding of its goal are
essential if the introduction of bias is to be
minimized. In general, this means that all
patients who started on the protocol should be
included in the data analysis. Multiple analyses
of a sample can introduce bias by summing type I
errors in the subgroups and necessitating
additional study to confirm results. The results
of subgroup analysis usually are given in the
context of the larger study. Typically, data are
presented in the form of Kaplan-Meier curves,
which represent probabilities of survival over
time. Responses to treatment also may be presented
in terms of objective measurement of tumor
response, such as shrinkage.
Phase III trials usually require
more elegant measures than phase I or II trials,
because they may involve more than one
randomization, windows, and stratification. Simple
traditional analyses involving 2-way comparisons
use chi-square. Additional computations on
subgroups are performed but must be considered in
light of the error potential of multiple group
analyses.
Phase IV trials
Phase IV trials apply positive
findings from research centers to generic use in
the community. These can include large-scale
population analysis for the purpose of marketing
and promotion by the company or for surveillance
as mandated by the FDA. Phase IV trials also can
be instituted for safety and efficacy analysis of
old drugs and can include the use of controlled
randomized studies.
Leukemias
Leukemias are the most common
type of childhood cancer, comprising 25% of new
diagnoses. The greatest advances in treatment have
occurred in leukemias, in no small part because of
the ability to treat relatively large numbers of
patients with uniform treatment plans that can be
evaluated.
Acute lymphoblastic leukemia
Nearly 80% of childhood
leukemias are ALL. The treatment of ALL has
yielded the first real success story of oncology.
Multi-component chemotherapy
regimens have resulted in long-term survival rates
approaching 90% in patients with favorable
prognostic factors, which include presenting
peripheral white blood cell count less than 20,000
X 103/mL; age older than 1 year and
younger than 10 years; early pre–B-cell phenotype;
presence of the TEL-AML1 translocation;
lack of mature T-cell, B-cell, and myeloid-cell
markers; lack of 9;22 translocation; early
remission; lack of CNS disease; female sex; lack
of mediastinal mass or organomegaly; initial
hemoglobin greater than 10 g/dL; platelet count
greater than 100 X 109/L; good
nutritional status; and normal immunoglobulin G (IgG)
levels.
The advent of modern molecular
techniques has resulted in the further dissection
of ALL into multiple subtypes with therapeutic
implications. For example, the recently described
TEL-AML1 translocation is present in
approximately 20% of pediatric cases of ALL. The
TEL-AML1 translocation now is considered
to be a favorable prognostic indicator for the
outcome of ALL, while the presence of the
so-called Philadelphia chromosome, a 9;22
translocation involving the bcr and
abl oncogenes, is a poor prognostic
indicator.
Acute myelogenous leukemia
An additional 18% of childhood
leukemias are AML. This ratio of ALL to AML holds
true throughout childhood except for a
predilection for AML in the neonatal period.
AML comprises a heterogeneous
array of subtypes termed M0 to M7.
Again, molecular diagnostic
methods have advanced the ability to subtype
myeloid leukemias; the analysis of translocations
is helping to define and confirm the histologic
designations. For example, the 8;21 translocation
associated with the M2 subtype is found in 15% of
patients with AML. Interestingly, this
translocation is a favorable predictor of
long-term survival. Similarly, the M3 subtype,
which is associated with a 15;17 translocation,
also has been correlated with a favorable outcome
by virtue of its response to therapy with all–trans-retinoic
acid. In contrast, the 9;11 translocation
associated with the M4 and M5 subtypes indicates a
poor prognosis. This abnormality is observed in
most individuals with AML following treatment with
etoposide.
Chronic leukemias
Chronic leukemias comprise fewer
than 5% of pediatric leukemias. Chronic
myelogenous leukemia (CML) is the most common type
and corresponds to the adult type of CML that is
marked by the Philadelphia chromosome. This adult
type of CML appears in older children (>4 y) and
is linked to radiation exposure in many
individuals with CML. Juvenile CML is a disease of
younger children, with most diagnosed in children
younger than 2 years. Other rare forms of chronic
childhood leukemia include myelomonocytic,
monocytic, and lymphocytic.
Brain Tumors
Tumors of the CNS constitute the
other major type of childhood cancer. A full 20%
of childhood cancers involve brain tumors.
Patients with CNS tumors remain an underreported
segment of the pediatric cancer population because
only half are referred to specialty centers.
Morbidity is clearly the greatest problem in brain
tumors, since a great many of these tumors are in
locations that are difficult to treat. Unlike
adult brain tumors, most true childhood brain
tumors occur in the posterior fossa.
Brain tumors are heterogeneous,
which makes their categorization a difficult
matter. The most common single entity brain tumor
in children is medulloblastoma, which comprises
10-20% of childhood brain tumors and 40% of those
in the posterior fossa. Most brain tumors involve
the posterior fossa, chiefly medulloblastomas and
glial tumors. Most CNS tumors are glial tumors,
which are classified by location as supratentorial,
cerebellar, or brainstem. Unique variants within
each of these groups have strong prognostic
significance. For example, patients with exophytic
gliomas do extremely well, while individuals with
diffuse infiltrative tumors do poorly.
Hodgkin Disease
Hodgkin disease, which comprises
5% of childhood cancers, peaks in children younger
than 14 years, young adults, and adults older than
55 years. Most statistical reports comment on
childhood cancers up to age 14 years. Thus, the
overall impact of Hodgkin disease in the
adolescent population tends to be understated.
Like non-Hodgkin lymphoma (NHL), Hodgkin disease
has been reported to be associated with
immunodeficiency and Epstein-Barr virus (EBV) as
well as cytomegalovirus and human herpesvirus 6.
Classification of Hodgkin
disease includes specific subtypes, including
nodular sclerosing, lymphocyte predominant, mixed
cellularity, and lymphocyte depleted. Nodular
sclerosing appears to be the most common subtype,
and lymphocyte depleted seems to be associated
with more severe disease and poorer outcome.
Hodgkin disease survivors remain at high risk for
secondary tumors, a phenomenon that may indicate
an underlying immunodeficient state. Breast cancer
in young patients with a history of Hodgkin
disease has been associated mostly with use of
radiation as a treatment modality.
Burkitt
Lymphoma
Burkitt lymphoma, a type of NHL,
is associated with EBV infection and is endemic on
the African continent. Burkitt lymphoma comprises
roughly one half of all incidents of NHL, which
translates to an incidence of approximately 2-3%
among childhood cancer. In its endemic form,
Burkitt lymphoma can occur at an incidence of up
to 50 times more frequently. Endemic Burkitt
lymphoma is associated with EBV and appears to
occur in equatorial Africa. Additional
environmental factors appear at work in the
pathogenesis of Burkitt lymphoma, as the endemic
form differs from even the sporadic form, which
also can be found along with EBV in North America
as the breakpoints of the 8:14 translocation
differ.
Small, Round, Blue-Cell Tumors
The predominant solid tumors in
children are the small, round, blue-cell tumors,
which together comprise approximately 30% of
childhood malignancies. Within this group, subtype
classification has been both obvious and muddled.
For instance, the primitive neuroectodermal tumors
remain a classification of great controversy,
resulting in differences in reporting and
treatment. Fortunately, most of the small, round,
blue-cell tumors have characteristics that lend
themselves to pathologic analysis. All of the
following tumor types are considered small, round,
blue-cell tumors.
Neuroblastoma
Neuroblastoma is the most common
non-CNS solid tumor; its round blue-cell
appearance is marked by neuropils on specimens
stained with hematoxylin and eosin. Both long-term
survival and short-term treatment remain
challenges in caring for patients with
neuroblastoma. Interestingly, age of presentation
has prognostic implications. The form that emerges
in infancy carries a much better chance of
long-term survival and is marked by a lack of
N-myc amplification; hyperdiploidy;
low-stage, limited distant sites in stage I or II
disease (<10% have marrow, liver, or skin
involvement); absence of chromosome arm 1p
abnormalities; and evidence of neuronal
differentiation. However, the form that emerges in
older children (ie, aged 1-10 y) has a much poorer
prognosis.
Non-Hodgkin lymphoma
Lymphomas make up a large, if
heterogeneous, category of childhood cancers.
Chief among these are the NHLs, which comprise 6%
of pediatric cancers. NHL is a disease of younger
children and overall has a predilection for males,
probably accounted for by the subset of T-cell
lymphomas, which occur predominantly in males. A
major factor in NHL is the association with
immunodeficient states secondary to underlying
genetic diseases, viral infection, or medicines.
Wilms tumor
Wilms tumor is the most common
renal tumor, constituting approximately 5-6% of
childhood cancers. As in neuroblastoma, age
impacts prognosis; presentation in infancy is
associated with better outcome. Wilms tumor is
associated strongly with a host of genetic
syndromes, including Beckwith-Wiedemann syndrome;
Wilms, aniridia, genitourinary abnormalities,
mental retardation (WAGR); Denys-Drash syndrome;
and Bloom syndrome. Studies of chromosome 11 have
led to the description of the WT1 and
WT2gene products, which are associated with
WAGR and Beckwith-Wiedemann, respectively.
Prognostic factors associated with long-term
survival include low-stage disease, favorable
histology, and young age.
Retinoblastoma
Retinoblastoma is the classic
tumor that led to the development of the "2-hit"
hypothesis of carcinogenesis. Study of family
trees and analysis of known mutations have
demonstrated a breakdown in incidence as
unilateral plus sporadic (60%), unilateral plus
inherited (15%), and bilateral plus inherited
(25%). Hereditary cancer occurs earlier and is
more likely to be bilateral, implying that a
second "hit" has occurred in more than one
location, the first "hit" already having been
inherited in the germline.
Incidents of sporadic cancer are
simply more likely to be unilateral by virtue of
the lower likelihood of 2 hits occurring in a
normal somatic cell. Inherited incidents of
retinoblastoma illustrate the importance of the Rb
protein product in the suppression of
tumorigenesis in that patients with inherited
retinoblastoma remain at risk for other tumors,
chiefly osteosarcoma.
Rhabdomyosarcoma
Rhabdomyosarcoma is another
solid tumor with an incidence that peaks in young
children (<6 y) and again in early adolescence.
This incidence roughly correlates with tumor type
in that younger patients generally are diagnosed
with head and neck tumors, and the histology is
usually embryonal, while older patients are more
likely to have tumors in the extremities with
alveolar histology. Generally, patients with
embryonal tumors and individuals with
hyperdiploidy have better outcomes; however, these
data remain somewhat controversial.
Osteosarcoma
Osteosarcoma is a bone tumor
associated with the rapid bony growth
characteristic of the adolescent growth spurt,
thus contrasting with Ewing sarcoma, which is a
bony tumor that is not associated with rapid bony
growth. Osteosarcoma is more common in patients
who are taller than their peers, and girls with
osteosarcoma are diagnosed at an earlier age than
boys. Tumors are localized to the growth plates of
long bones. Radiation and alkylating agents have
been implicated in the etiology of osteosarcoma
along with retinoblastoma and Li-Fraumeni
syndrome. Osteosarcoma subtype is probably the
most important prognostic factor.
Well-differentiated variants (eg, parosteal and
intraosseus tumors, diploid tumors) are associated
with better outcomes.
Ewing sarcoma
Ewing sarcoma is a collection of
tumors that includes peripheral primitive
neuroectodermal tumors and primary bony tumors.
The diagnostic standard involves the presence of
either the 11;22 or the 21;22 translocation, at
least one of which is found in as many as 95% of
individuals with Ewing sarcoma. An interesting
feature of Ewing sarcoma is its extreme rarity
among African Americans. Although most incidents
are found in the second decade of life, Ewing
sarcoma occurs more throughout the age spectrum
than osteosarcoma; Ewing sarcoma is not associated
with rapid bone growth and may be found anywhere
along bone, adjacent soft tissue, or even as an
isolated soft tissue mass.
In general, pediatric cancer
causation has been a contentious issue, and
relatively few causative factors have been
identified. Certainly, the greater numbers of
adults with cancer have proven the ability to
ascertain causative factors, such as alcohol and
smoking, whereas the small numbers of children
with cancer have made environmental causation a
much more difficult area to evaluate. However,
analysis for inherited factors has been
increasingly fruitful and is expanding in scope,
given the explosion in availability of molecular
biologic technology and resources engendered by
the Human Genome Sequencing Project.
Inherited Predisposition
At its most basic level, cancer
is a genetic disease. Production of genetic
instability that confers some kind of mutator
phenotype is most likely the chief characteristic
of any inherited cancer predisposition. These take
one of the following forms: (1) mutations in key
genes that are involved directly in tumoral
development (eg, WT1, WT2), (2) mutations
in genes that generate mutations and gross
chromosomal deletions in key loci (eg, in Fanconi
anemia and mismatch repair), (3) mutations in
genes that are involved directly in DNA repair of
specific lesions (eg, xeroderma pigmentosum), and
(4) complex chromosomal syndromes that incur
cancer susceptibility.
Down syndrome
Children affected with Down
syndrome have a 1% risk of developing leukemia
before they are aged 10 years; the ratio of types
is different than in children overall, in that 60%
of children with Down syndrome develop ALL and 40%
develop AML. Generally, children with Down
syndrome and ALL appear to have a worse prognosis
than children with ALL. In contrast, children with
Down syndrome and AML tend to have better
outcomes. This may reflect the association of Down
syndrome with transient myeloproliferative disease
of infancy, which not only resembles congenital
leukemia but also confers a 30% risk of subsequent
AML.
Turner syndrome mosaicism or
androgen insensitivity syndrome
Retention of the Y chromosome in
females with Turner syndrome mosaicism or in
androgen insensitivity syndrome results in an
increased lifetime risk of gonadoblastoma. This
risk is as high as 25% by adulthood.
Wilms tumor
Association of gross deletions
at the 11p13 locus with Wilms tumor led to
isolation of the WT1 gene. Clinical
abnormalities associated with WT1
mutations include aniridia, genital abnormalities,
and mental retardation. As many as 40% of
individuals with Wilms tumor have been reported to
have some familial component.
Increased growth syndromes have
been associated with Wilms tumor, including the
Beckwith-Wiedemann syndrome and hemihypertrophy.
Beckwith-Wiedemann is linked to chromosome band
11p15, where a putative WT2 gene resides;
insulin growth factor 2 and p57kip2 are
the leading candidates for the WT2 tumor
suppressor gene.
Mendelian
Inheritance of Genetic Cancer Predisposition
Autosomal dominant disorders
Knudson was studying
retinoblastoma, an autosomal dominant disorder,
when he first described the "2-hit" hypothesis of
carcinogenesis. This hypothesis describes the
process whereby, given the transmission of these
disorders genetically through the germline, the
loss of a second allele in a predisposed patient
leads to the onset of cancer at an earlier age.
These disorders are more likely than other cancers
to be associated with bilateral and multiple
tumors. Concomitant with this risk is the risk of
multiple tumors at various times during the
lifetime, depending on the tissue at risk.
Retinoblastoma: The deleted
Rb gene confers not only increased risk to
the patient born with the mutation but also
entails unknown risk for 2 other groups: patients
with newly diagnosed sporadic cases and the
familial carriers who do not develop
retinoblastoma as children. Mutation in the Rb
gene confers a lifetime risk of osteosarcoma and
melanoma.
Li-Fraumeni syndrome: p53
represents the most commonly mutated gene in human
cancers and is the responsible dysfunctional gene
in the rare familial Li-Fraumeni cancer syndrome (LFS).
Numerous cancers cluster in LFS, including
sarcomas, breast cancer, leukemia, brain tumors,
and adrenocortical carcinoma. The study of LFS has
led to greater understanding of cancer in general,
as p53 appears to be a convergence point
of many cancers in the long progression of the
multistep process of carcinogenesis.
Familial colon cancer: The
development of multiple colonic polyps has been
associated with the early development of colon
cancer and hepatoblastoma. The APC gene
was found by positional cloning and affects
signaling through the beta-catenin pathway.
Hereditary nonpolyposis colon
cancer (HNPCC): This first was defined as a
genomic instability disorder in which the
underlying genetic defect promoted the loss of the
other allele, giving rise to the tumor. The HNPCC
group involves at least the mismatch repair
proteins that are implicated in a whole array of
adult cancers. Analysis of the mismatch repair
genes is used at the protein level and as an in
vitro test for carrier status for HNPCC.
Multiple endocrine neoplasia
(MEN): The MEN gene complex is marked by an
association of cancers of the thyroid,
parathyroid, pancreas, pituitary, and adrenal
medulla. Pathogenesis of the MEN type 2 syndrome
appears to be due to activating mutations of the
ret oncogene rather than to a 2-hit
mechanism.
Neurofibromatosis:
Neurofibromatosis type 1 (NF-1) is one of the most
common genetic syndromes and is marked by a
propensity to brain tumors and peripheral nerve
sheath tumors. Defects in ras GTPase,
termed neurofibromin, are sporadic in at least one
half of the cases of neurofibromatosis that can be
detected in the general population. The frequency
is 1 in 3000 persons. Patients with NF-1 are prone
to optic gliomas, most commonly in early
childhood, along with a risk of gliomas in other
locations. A link to development of myeloid
leukemias also has been described, which is
consistent with the connection between ras
mutations and myeloid disease. Association with
numerous other diseases has been reported but not
proven.
Tuberous sclerosis: A syndrome
of seizures, mental retardation, and angiofibromas,
tuberous sclerosis is associated with a range of
benign growths. Cardiac rhabdomyomas are a problem
of infancy, while retinal hamartomas and giant
cell astrocytomas develop later in childhood.
von Hippel-Lindau syndrome:
This syndrome involves an association of renal
cell carcinoma, retinal and cerebellar angiomata,
and pheochromocytoma. The VHL gene
product is an elongin that is responsible for
normal transcription completion.
Autosomal recessive disorders
Xeroderma pigmentosum: This
disorder results from several genetic
complementation groups that are part of the
nucleotide excision repair system and
transcriptional apparatus. Patients with xeroderma
pigmentosum have an increased risk of basal cell
carcinoma, squamous cell carcinoma, and melanoma.
Neurologic and other skin findings are also part
of the related disorders trichothiodystrophy and
Cockayne syndrome.
Ataxia telangiectasia: This
radiation hypersensitivity syndrome comprises a
constellation of ataxia, oculocutaneous
telangiectasia, and increased incidence of
lymphoid malignancies. The gene product
responsible for this disease is the ATM PI-3
kinase, which participates in the rad50-BRCA1
epistasis group and probably is involved in
double-strand break repair by homologous
recombination.
Fanconi anemia (FA): This
disorder of hypersensitivity to bifunctional
alkylating agents is marked by congenital defects,
bone marrow failure, and multiple cancer
susceptibility, most commonly AML. FA has at least
8 genes that are defective among its known
complementation groups. Although many have been
cloned, the molecular basis for FA remains
elusive.
Immunodeficiency States
Although evidence is not
plentiful, immune surveillance clearly plays a
major role in tumor prevention. The most dramatic
example is CML, in which a measurable
graft-versus-leukemia effect occurs whereby
immunosuppression to avoid graft-versus-host
disease results in a decrease in leukemia-free
survival. In addition, theoretical analysis of
tumor kill after chemotherapy as well as
measurement of residual disease both demonstrate
that the tumor is still present after therapy.
Reliance upon the host’s immune system is assumed
to clear disease. Thus, logically, immunodeficient
states can be postulated to engender cancer
susceptibility.
Severe combined
immunodeficiency
Patients with severe combined
immunodeficiency are difficult to assess because
of the severity of their underlying defect;
however, an inherent propensity toward lymphoid
malignancy is clear. Those patients who live
longer may have some residual immune system and
thus a longer time for cancer development.
Wiskott-Aldrich syndrome
This immunodeficiency disorder
is characterized by thrombocytopenia, eczema, and
T-cell dysfunction and carries an increased risk
of NHL.
Lymphoproliferative syndromes
These syndromes, which may be
both genetic and related to therapy, confer an
increased risk of lymphoid proliferation triggered
by EBV infection. In the X-linked form of the
disease, EBV infection accounts for 70% of deaths.
After prolonged immunosuppression (eg, chronic
graft-versus-host disease following bone marrow
transplantation) an increased susceptibility to
lymphoproliferative disease occurs.
Human immunodeficiency virus
HIV has not left the pediatric
population unaffected, in spite of promising
regimens for preventing vertical transmission and
promotion of safe sex practices. Children
generally follow a more rapid progression to AIDS.
The spectrum of cancers associated with HIV
includes Kaposi sarcoma, NHL (especially CNS), and
leiomyosarcoma.
