TABLE OF CONTENTS
INTRODUCTION
REGISTRY OVERVIEW
HOW TO USE THIS REPORT 2
Overview
METHODS
Calculation of Rates
Standard Error of Rates 4
Confidence intervals and Significance Testing
Graphic Presentations: Pie &, Bar Charts and Maps . . . . . 6
DISCUSSION OF RESULTS
GRAPHS & DETAILED DATA TABLES 9
Statewide Pediatric Cancer Incidence Rates, by Race, for Detailed
Diagnostic Groups 13
Statewide Cancer Incidence & Mortality Rates, by Race & Sex, for
Each Major Pediatric Site
Pediatric Cancer Incidence & Mortality,
Compared with National Data
County Incidence Rate Maps for Selected Major Pediatric Cancers . . 43
Detailed County Incidence Rate Tables for Selected Major Pediatric
Cancers
57
NEW JERSEY VS SEER ICD SITE CODE COMPARISON 68 """""'
REFERENCES 70
(page break)INTRODUCTION
This report presents statewide and county specific age-adjusted incidence and mortality rates for
childhood cancer in New Jersey during the period 1980-1988 as well as comparative national
data. It focuses on total cancer and on the several leading cancer sites that account for most
pediatric cancer incidence. We have used the terms childhood and pediatric interchangeably, to
mean the ages from birth through the first fourteen years of life.
The primary goal of this report is to provide a concise volume of state, county and national level
pediatric cancer data to health planners and researchers who are working on a state, regional or
local scale in New Jersey. Data are provided statewide for four population subgroups, white
males, white females, black males and black females, which together account for 98% of the
New Jersey population. Rates are also provided for broader population groups, i.e. all males, all
females, all Whites, all Blacks, and total population. National data were published only on white
children, and we have kept this convention for the national comparisons. At the county specific
level, lower numbers of cases required that we examine all children regardless of race or gender.
For information regarding adult cancer rates in NJ, please refer to The Most Frequent Cancers in
New Jersey: Incidence, Stage, Survival and Mortality Rates 1979-1985 & Addendum1,2. The
Most Frequent Cancers in New Jersey: Incidence and Mortality Rates by County, Race& Sex
1979-19853 and Cancer in New Jersey: Incidence and Morraiity Rates by Race and Sex for
Total Cancer & Eleven Major sires 1986-19884. REGISTRY OVERVIEW
The New Jersey State Cancer Registry (NJSCR) was established by legislation (NJSA 26:2-104
et.seq.) as a population-based incidence registry and includes all cancer cases diagnosed
among New Jersey residents since October 1, 1978. The NJSCR serves the entire State of New
Jersey, which includes a population of approximately 77 million people. For the years 1980
through 1988, an average of 229 incident cases of cancer were reported each year among New
Jersey children, aged fourteen or younger, representing less than one percent of the total
number of cancers reported among all age groups. Over the same period there were an average
of 61 deaths among children whose primary cause of death was recorded as cancer each year.
New Jersey regulations (NJAC 8:57A) require the reporting of all newly diagnosed cancer cases
to the NJSCR within three months of hospital discharge or six months of diagnosis (whichever is
sooner). All primary malignant and in-sim neoplasms, except certain carcinomas of the skin, are
reportable to the NJSCR. Reports are filed by hospitals, diagnosing physicians, dentists, and
independent clinical laboratories. In addition, reporting agreements are maintained with
neighboring states so that New Jersey residents diagnosed with cancer in health care facilities
outside the state can be identified.
The information collected by the NJSCR includes basic demographic characteristics of the
patient, medical information on each cancer diagnosis (such as the anatomic site, histologic type
and summary stage of disease), and annual follow-up status (alive/dead). These data along with
underlying cause of
(page break)1
death (where applicable) are incorporated into the basic data set. Medical researchers and
health planners find the data of the NJSCR to be a valuable source of information for a multitude
of study and planning purposes.
HOW TO USE THIS REPORT Overview
This report titled Childhood Cancer in New Jersey: Incidence and Mortality Rafes by Race, Sex
& Countyfor Total Cancer and Major Shes 198~+1988, presents county level incidence rates for
total cancer and for cancer sites occurring most frequently among children residing in New
Jersey. Although intended to stand alone as a reference document on pediatric cancers in New
Jersey, it may be used in conjunction with our prior publications as referenced above. Because
this pediatric report was designed as a reference document, very little interpretation of results
has been provided. Comparison of New Jersey rates among race-sex groups to national data' is
provided. Some of the national incidence data were reanalyzed from the National Cancer
Institute Public Use Tape in order to calculate confidence intervals and perform significance
testing between New Jersey and Surveillance Epidemiology and End Results (SEER) incidence
rate estimates. In addition, comparisons with national mortality data for the same time period
were also made.
Interpretation of the data should be approached after taking the following factors into
consideration: (1) The presentation of data has been limited to the most common primaly sites
among children in New Jersey. (2) Categorization of individuals by primary site, gender, and
race often leaves very small numbers in the cells used for calculating incidence and mortality
rates, which leads to wide confidence intervals. (3) Pediatric cancer is considered a rare
disease. (4) Small differences in the numbers of cases diagnosed in any given year can cause
substantial variation in rates from year to year. (5) The presentation of a rate for each three year
period as well as the nine year rate (an overall rate) minimizes some of this yearly fluctuation.
Another factor to consider when interpreting the data in this report is the reliability of the
population estimates used in the rate denominators. The population estimates used in this report
were obtained from the National Cancer Institute (NCI). These data are an approximation of the
population distribution in New Jersey derived by the interpolation of figures between the 1980
and 1990 Census estimates and are thus subject to error.
The effects of migration must also be kept in mind when examining geographic comparisons and
interpreting the data. The rates presented here are based on the patient's place of residence at
diagnosis (for incidence rates) or death (for mortality rates), and do not account for individuals
changing residence just prior to diagnosis or death. A long latency period is associated with the
development of most adult cancers. Although a long latency period is less prominent a factor
when examining pediatric cancers, migration effects must be considered, and confounding
exposures cannot be ruleil out. Residence at the time of diagnosis or death may not have any
relevance to initiating or promoting exposure(s).
(page break 2)
The quality of the NJSCR data has improved over time. In 1979, 94% of the cases were
microscopically confirmed, and 15% of the cases were identified solely from death certificates.
By 1988, 97% of the cases were microscopically confirmed, and less than 6% were diagnosed
exclusively by death certificate. Although some of these "death certificate only" cases may
possibly have been diagnosed prior to our reference date of October 1, 1978, death certificate
only cases have been included in the data presented in this report.
This report begins with pie and bar charts which summarize the cancer proftle of New Jersey
children, along with a table which breaks down total cancer into a detailed list of types and
subtypes for each of the white and black races. This is followed by a statewide race and gender
comparison for each of the major childhood cancers arranged in alphabetic order by major site
group. Next, tables showing incidence and mortality among white New Jersey children are
compared with the available national data from NCI-SEER reports. Finally, incidence data for
each county in New Jersey are contrasted with the state rates, using maps and detailed data
tables.
METHODS
Cancer incidence data were obtained from the NJSCR, while cancer mortality data were
obtained from the Vital Statistics Program of the New Jersey State Department of Health.
Cnkulolion of Roles
A cancer incidence (or mortality) rate can be defined as the number of new cases of cancer (or
deaths attributed to cancer) detected during a specified time period in a specific population
These rates are most commonly expressed as cases or deaths per 100,000 person-years of
observation Cancer occurs at different rates in different age groups, and population subgroups
defined by gender, race, and county have different age distributions. Therefore, before a valid
comparison can be made between rates, it is necessary not only to adjust the rates by age but
also to standardize the rates to the age distribution of a standard population. In this report, the
1970 US Standard Population was used to maintain comparability to NCI-SEER publications.
The first step in this procedure was to determine the agespecific rates. For each five year age-
group (within each racesex group, far each county or for the entire state), the following formula
was applied:
r(sub)a= d(sub)a/(t x P(sub)a) x 100,000
where:
r(sub)a = the age-specific rate for agegroup a,
d(sub)a = the number of events (cancer diagnoses or deaths) in the age-group during the
time interval,
(page break 3)
t = the length of the time interval (in years), and
P, = average size of the population in the age-group during the time interval (mid- year
population or average of mid-year population sizes).
The age-specific, race and sex specific population estimates for each county were provided by
the SEER Program of NCI through an interagency agreement with the US Bureau of the Census.
The age-specifre population data and age-specific rates are not shown in this report but are
available from the NISCR upon request.
In order to determine the age-adjusted and standardized rate, a weighted average of the age-
specific rates was calculated, using the age distribution of the standard population to derive the
age-specific weighting factors (Rothman, 1985)6. This is the technique of direct standardization,
which uses the following formula.
(Epsilon) i. x SCd. P.
R="
x 100,000
Epsilon 5f d. P.
a-i where:
R = the age-adjusted rate
r, = the age-specific rate for age-group a, and
Std.P, = the size of the standard population in each age group a. While age
adjustment and standardization facilitates the comparison of rates among different
populations, there can be important age-specific differences in disease occurrence or
death which are not apparent in comparisons of the age adjusted rates (Breslow and Day)'.
Moreover, age is only one of a number of known risk factors that might vary
systematically and account for differences in cancer incidence and mortality among
populations of interest.
To facilitate comparison of county rates with the state rate in this publication, an "overall" rate
was calculated. To obtain the overall rate, the same calculations that were applied to each
calendar year of observation were applied to the entire nine year period. This was done rather
than an average of nine individual yearly rates, so that an estimate of the standard error could be
used to calculate the confidence interval around the point estimate.
Stnndard Error of Roles
The calculation of the standard error for an incidence or mortality rate serves to highlight the fact
that disease occurrence is subject to fluctuation over time, and that a measurement made at any
particular point in time serves only to estimate the true "underlying" rate of disease occurrence or
death in a given population. In particular, it should be noted that rates in smaller populations are
likely to show more variability over time than those in larger populations, and therefore rates
calculated during the same time period are inherently less reliable for smaller populations than
for larger ones. This is an important consideration when comparing measures of incidence and
mortality among geographically
(page break 4)defined units such as the counties of New lersey, where there is a ten-fold range
in population size.
Standard errors in this report were calculated using a modification of the method of Chiang
(1961), which assumes that age-specific rates are independently, binomially distributed.8 The
formula used herein was:
where:
r, = the age-specific rate in age group a
t = the length of the time interval (in years)
P, = the average age-specific population in age group a during the time interval (mid-year
population or average of mid-year populations)
w, = the proportion of the standard population in age group a, and
n = the number of age groups (here, three fiveyear age groups were used). Confidence
Interviews and Significance Testing
Confidence intervals provide a concise way to visualize the reliability of the point estimates for
the age-adjusted county. state and national rates. The confidence interval is defined as the
range between the upper and lower confidence limits. A narrower range shows a greater
confidence in the point estimate, while a wider range indicates less surety. The confidence
intervals were chosen to correspond exactly to statistical significance testing at a p<0.05 level
The estimates for each confidence limit (which would correspond to error bars on either side of
the point estimate for each age-adjusted rate) were calculated by:
Confidence Limits = R +/- (1.96 x S.E. )
where:
R = the age-adjusted rate for the race-sex group, and
SE = the standard error of the point estimate, calculated as previously derailed under
Standard Error of Rates.
Numerically, confrdence intervals are presented as a pair of numbers representing the potential
range within which the true rate is expected to fall. Therefore, the confidence interval (0.5,1.6)
would be literally interpreted as "We are 95% confident that the true rate lies in the range 0.5 to
1.6." In
(page break 5)
cases where the above formula would have resulted in a ]ower limit below zero, a value of zero
is given for the lower limit, since a negative incidence (or mortality) rate is impossible When a
confidence interval includes zero, there is the possibility that the true rate in the population might
be null. This is seen to occur most often when a rate is low, or when low numbers of cases or a
small population are involved. By using confidence intervals, it is possible to compare these
rates for two or more counties, to compare a county rate with the state rate or the state to the
nation. When the confidence interval for one rate overlaps that of another, there cannot be
certainty that the rates are truly different.
The 9546 confidence intervals corresponding to the range of values which fell within 1.96
standard errors of the respective rate were determined for each overall nine year incidence (or
mortality) rate. Tests of statistical significance were conducted involving the comparison of rates
for the state versus the appropriate national data (SEER incidence or United States mortality)
and for each county against the corresponding statewide rates. These tests were conducted by
examining whether the 9546 confidence intervals for the county-level rates overlapped with the
95% confidence intervals for the corresponding statewide rates.
Graphic Presentations: Pie & Bar Charts and Maps
Pie charts are presented showing the distribution of each type of cancer among children and the
demographic differences in these distributions among Blacks, Whites, males and females. A bar
chart shows the actual number of cases contributed by each demographic group, for each type of
cancer. Time trend graphs are presented to show the relative relationships and stability of New
Jersey pediatric incidence and mortality rates with the corresponding national data (SEER
incidence and United States mortality rates, respectively).
Maps are used to examine variations in cancer incidence and mortality rates at a regional level,
i.e., focusing on groups of counties, with the understanding that the relative instability of rates in
counties with smaller populations or population subgroups may confound regional patterns. For
each of the selected cancers, both the county specific rates and the results of the statistical
significance testing were imported into a commercial mapping program. A map presenting the
total range of rates divided into five equally spaced ranges indicating the magnitude of each
county's incidence rate has been included. For total cancer and cancer of the brain and central
nervous system, a second map displaying the results of the significance testing indicates
whether the corresponding county rate was significantly higher, not significantly different, or
significantly lower than the state rate for that cancer. For all other sites, no statistically significant
variation on the county level was observed.
DISCUSSION OF RESULTS
New Jersey's statewide pediatric cancer incidence and mortality rates are similar to national
rates and follow the same trends. However, for the period 1980-1988, New Jersey pediatric
incidence rates for total cancer and soft tissues were significantly higher than the corresponding
SEER rates.
(page break 6)
Moreover, New Jersey pediatric mortality rates were not significantly higher than the
corresponding United States mortality rates.
Approximately 53% of all pediatric cancer in New Jersey is due to leukemia or cancers of the
brain and central nervhus system. Both of these types of cancer showed higher annual age-
adjusted rates during the years 1981 through 1984, peaking in 1983, before converging with the
corresponding SEER incidence rates. It is not possible to determine a cause for this transitory
increase from data collected by the NJSCR. SEER relative survival data indicate that 80% of
white children diagnosed with brain cancer will survive one year, decreasing to 57% by five
years. In recent years, a substantial increase in survival has been observed for acute
lymphocytic leukemia (A.L.L., the most frequently occurring type of childhood leukemia),
Hodgkin's Disease, non-Hodgkin's lymphoma and soft tissue cancers.'
Examination of New Jersey incidence data by race and gender indicates that white children have
a 28% higher risk of developing cancer than do black children, and male children have a 17%
higher risk than do female children. Mortality data show that male children face an 18% higher
risk of death due to cancer than female children, however no significant differences are seen
between the rates for white and black children for the years 1980-1988.
(page break 7)
(page break 8)
GRAPHS AND DETAILED DATA TABLES