America’s Children and the Environment Health Methods Childhood Cancer Health Childhood Cancer Draft

Health: Childhood Cancer

Methods for Childhood Cancer

Indicator H4 Cancer incidence and mortality for children ages 0 to 19 years,
1992-2018.

See below for Indicator H5 Cancer incidence for children ages 0 to 19 years, by type, 1993-2018.
Summary

Since 1973, the Surveillance, Epidemiology, and End Results (SEER) Program of the National
Cancer Institute (NCI) has been collecting and publishing cancer incidence data from population-
based cancer registries that currently cover a total of 18 geographical areas in the United States
and one quarter of the population. Since the coverage area has expanded over time, the trend
analyses for indicator H4 were based on the SEER 13 registries that cover 13.4% of the U.S.
population based on 2010 Census data. Data include the type of cancer, age at diagnosis, year of
diagnosis, sex, race, and ethnicity. For indicator H4, cancer incidence rates for malignant cancers
in children ages 0 to 19 years were calculated using the SEER* Stat software provided by the
NCI. SEER* Stat calculates incidence rates for each age group by dividing the number of new
cancer cases in a given year by the total population for that age group in the cancer registry
geographical area in that year. SEER*Stat then calculates age-adjusted incidence rates as a
weighted average of the rates for each five-year age group, where the weights are the proportions
of persons in each age group in the selected standard population; for these analyses the incidence
rates were age-adjusted to the 2000 U.S. Standard Population.

National cancer mortality rates were also obtained from SEER, which uses data from the
National Vital Statistics System, administered by the National Center for Health Statistics. Data
include the type of cancer, age at death, year of death, sex, race, and ethnicity. For indicator H4,
cancer mortality rates for malignant cancers in children ages 0 to 19 years were also calculated
using SEER* Stat. The mortality rate is calculated by dividing the number of deaths attributed to
a malignant cancer in a given year by the total population for that age group in the cancer
registry geographical area in that year. The mortality rates were age-adjusted to the 2000 U.S.
Standard Population.

Supplementary Tables: Tables H4a, H4b, and H4c provide the age-adjusted cancer incidence and
mortality rates for children 0 to 19 years, stratified by race/ethnicity, sex, and age for 2016-2018.

DRAFT America's Children and the Environment
December 2021

Page 1

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Data Summary

Indicator H4. Cancer incidence and mortality for children ages 0 to 19 years.

Data

Cancer incic

ience of malignant cancers in SEER 13 registries for children ages 0 to 19 years.
J.S. mortality from malignant cancers in children ages 0 to 19 years.

Years

1992

1993

1994

1995

1996

1997

1998

Cancer
incidence

1,654

1,723

1,654

1,684

1,775

1,714

1,833

Cancer
mortality

2,417

2,419

2,354

2,275

2,224

2,261

2,176

Years

1999

2000

2001

2002

2003

2004

2005

Cancer
incidence

1,779

1,838

1,912

1,982

1,807

1,934

2,029

Cancer
mortality

2,243

2,271

2,226

2,271

2,233

2,223

2,183

Years

2006

2007

2008

2009

2010

2011

2012

Cancer
incidence

1,837

2,007

2,036

2,103

2,116

2,093

2,061

Cancer
mortality

2,035

2,068

2,039

1,964

1,925

1,922

1,995

Years

2013

2014

2015

2016

2017

2018



Cancer
incidence

2,043

2,159

2,229

2,143

2,054

2,047



Cancer
mortality

1,914

1,837

1,855

1,920

1,797

1,841



DRAFT America's Children and the Environment
December 2021

Page 2

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Calculation of Indicator

All the calculations for this indicator were carried out using SEER*Stat software, version 8.3.9.2.
In this section we detail the menu options required for carrying out these analyses.

Note that any User-Defined variable previously created using the File and Dictionary menus in
the same or any earlier session will not need to be recreated.

Incidence

Click the Rate Session button "E".

Data tab. Select "Incidence - SEER Research Data, 13 Registries, Nov 2020 Sub (1992-2018)."
For "Age Variable" select "Age recode with <1 year olds."

Statistic tab. Check "Rates (Age - Adjusted)" and "Show Standard Errors and Confidence
Intervals." For "Standard Population" select "2000 US Std Population (19 age groups - Census
P25-1130)." For "Age Variable" select "Age recode with <1 year olds."

Table tab. From the "Available Variables" panel, expand "Race, Sex, Year Dx," and select "Year
of diagnosis." Click the Row button. This tabulates the rates by calendar year.

Output tab. For "Display Rates as Cases Per" select "1,000,000." For "Number of Decimal
Places for" select "0.001."

Click the Execute button to run the analyses and create the output Rate Matrix.

Use the Matrix and Export menus to output the data to a text file.

Mortality

Click the Rate Session button "E".

Data tab. Select "Mortality - All COD, Aggregated With State, Total U.S. (1969-2018)
-------
Health: Childhood Cancer

Selection tab. For "Age at Death" click the Edit button and then select the age group 0-19 using
"Age recode with <1 year olds" = 00 years, 01-04 years, 05-09 years, 10-14 years, and 15-19
years. For "Other (Case Files)" click the Edit button and then follow the menus to select "Site
and Morphology," "Cause of death recode," and "All Malignant Cancers." This restricts the data
to malignant cancers in children ages 0 to 19 years.

Table tab. From the "Available Variables" panel, expand "Race, Sex, Year Dth, State, Registry"
and select "Year of Death." Click the Row button. This tabulates the rates by calendar year.

Output tab. For "Display Rates as Cases Per" select "1,000,000." For "Number of Decimal
Places for" select "0.001."

Click the Execute button to run the analyses and create the output Rate Matrix.

Use the Matrix and Export menus to output the data to a text file.

Age adjustment

For indicator H4, all cancer incidence and mortality rates were age-adjusted to the 2000 U.S.
Standard Population using 5 age groups for children. The calculations were carried out
automatically by the SEER*Stat software. The age-adjusted rates are often preferred when
comparing different populations (such as for different calendar years or different race groups)
since differences in the age-adjusted cancer rates are mainly attributable to factors other than
age. In general, very young children and adolescents will tend to have a higher crude
(unadjusted) cancer rate than children ages 5 to 14. The age adjustment replaces the distribution
of ages in each given population by the same standard age distribution, in this case the age
distribution of the U.S. population in 2000.

The calculation is shown by the following example which gives the age-adjusted cancer
incidence rate for 2018 for children ages 0 to 19 years.

Age-adjusted Cancer Incidence for 2018.

Age
Group [1]

Cases

12]

SEER 13
Population [3]

Crude Rate
(Cases per
Million) [4]
= [2]/[3]
x1,000,000

U.S. 2000
Standard
Population
[5]

U.S. 2000
Proportion
of Age
Group
0-19 [6]

Adjusted

Rate
(Cases per
Million) =
[4]x[6]

00 years

126

513,509

245.371

3,794,901

0.048

11.819

01-04 years

468

2,148,693

217.807

15,191,619

0.193

42.000

05-09 years

358

2,718,074

131.711

19,919,840

0.253

33.303

DRAFT America's Children and the Environment
December 2021

Page 4

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Age
Group [1]

Cases

12]

SEER 13
Population [3]

Crude Rate
(Cases per
Million) [4]
= [2]/[3]
*1,000,000

U.S. 2000
Standard
Population
[5]

U.S. 2000
Proportion
of Age
Group
0-19 [6]

Adjusted

Rate
(Cases per
Million) =
141x161

10-14 years

446

2,789,246

159.900

20,056,779

0.255

40.708

15-19 years

649

2,775,691

233.816

19,819,518

0.252

58.821

00-19 years

2,047

10,945,213

187.022

78,782,657

1.000

186.651

The first three columns give the number of new malignant cancer cases in 2018 and the total
population for each children's age group within the counties included in the SEER 13 cancer
registries. The number of cases divided by the SEER 13 population and multiplied by 1 million
gives the crude rate shown in the fourth column. For example, for the age group 00 years, there
were 126 cases in a population of 513,509 giving a crude rate of (126/513509) x 1000000 =
245.371 cases per million. The crude rate for all children ages 0 to 19 years is 187.022 (bottom
row).

The U.S. 2000 Standard Population for children is shown in the fifth column. This is the age
distribution of children for the entire United States in 2000, in five age groups (0, 1-4, 5-9, 10-
14, and 15-19). For analyzing children's populations only five age groups are needed (0, 1-4, 5-
9, 10-14, and 15-19). However, for analyzing children and adults the U.S. 2000 Standard
Population has 19 age groups, since the U.S. 2000 Standard Population also includes five-year
age groups up to 80-84 and the 85+ age group. The standard age distribution for children ages 0
to 19 years is shown in the sixth column. The U.S. 2000 Standard Population in each age group
is divided by the total U.S. 2000 Standard Population for children ages 0 to 19 years, i.e.,
78,782,657.

The age-adjusted population is a weighted average of the crude rates (column 4) weighted by the
U.S. 2000 Standard Population proportions (column 6). Thus, the first five rows of column 7
contain the products of columns 4 and 6, and the age-adjusted rate is given by the total for
column 7, shown in the bottom row as 186.651.

For this example, the crude rate for children ages 0 to 19 years, 187.022, is very close to the age-
adjusted rate, 186,651, reflecting the fact that in 2018 the age distribution for children 0 to 19
years in the SEER 13 regions closely matches the 2000 U.S. population age distribution. For
other years or for different race/ethnicity/sex groups, crude and age-adjusted rates can differ by a
much greater percentage.

Relative Standard Error

The uncertainties of the incidence and mortality rates were computed by the SEER* Stat software
under the assumption that the counts have Poisson distributions. The relative standard error is the
standard error of the rate divided by the estimated rate:

DRAFT America's Children and the Environment
December 2021

Page 5

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Relative Error (%) = [Standard Error (Rate) / Rate] x 100%

Rates with a relative error less than 30% were treated as being reliable and were tabulated. For
indicator H4, as well as the tables H4a, H4b, and H4c, described below, all the relative standard
errors were at most 30% and so all values were tabulated.

Rates by Race, Ethnicity, Sex, and Age

Supplementary Tables: In Tables H4a, H4b, and H4c, cancer incidence and mortality rates are
tabulated by race, ethnicity, sex, and age.

The following SEER*Stat data files were used for the analyses in Tables H4a, H4b, and H4c:

• Surveillance, Epidemiology, and End Results (SEER) Program (www.seer.cancer.gov)
SEER*Stat Database: Incidence - SEER Research Data, 13 Registries, Nov 2020 Sub
(1992-2018) - Linked To County Attributes - Time Dependent (1990-2018)
Income/Rurality, 1969-2019 Counties, National Cancer Institute, DCCPS, Surveillance
Research Program, released April 2021, based on the November 2020 submission.
Incidence data from the SEER 13 registries: Connecticut; Hawaii; Iowa; New Mexico;
Utah; Atlanta, Georgia; Detroit, Michigan; San Francisco-Oakland, California; Seattle-
Puget Sound, Washington; Los Angeles, California; San Jose-Monterey, California;
Rural Georgia; and the Alaska Native Tumor Registry. Downloaded November 2021.

• Surveillance, Epidemiology, and End Results (SEER) Program (www.seer.cancer.gov)
SEER*Stat Database: Mortality - All COD, Aggregated With State, Total U.S. (1990-
2018)
-------
Health: Childhood Cancer

Selection tab. Check "Malignant Behavior." For "Age at Diagnosis" click the Edit button and
then select the age group 0-19 using "Age recode with <1 year olds" = 00 years, 01-04 years,
05-09 years, 10-14 years, and 15-19 years. For "Race, Sex, Year Dx, Registry, County" click the
Edit button and select the years 2016-2018 using "Year of Diagnosis" = 2016, 2017, and 2018.
This restricts the data to malignant cancers in children ages 0 to 19 years diagnosed in 2016-
2018.

Table tab. From the "Available Variables" panel, expand "Race, Sex, Year Dx" and select "Sex."
Click the Row button. Use the menus "File," "Dictionary," and "Merge" to create a User-
Defined Merged variable Raceeth from the Race recode (W, B, AI, API) and Origin recode
NHIA (Hispanic, Non-Hisp) variables in the database:

• Raceeth = All if Race recode = White, Black, American Indian/Alaskan Native, Asian or
Pacific Islander, or Unknown

• Raceeth = WhiteNH if Race recode = White and Origin recode NHIA = Non-Spanish-
Hispanic-Latino

• Raceeth = BlackNH if Race recode = Black and Origin recode NHIA = Non-Spanish-
Hispanic-Latino

• Raceeth = AIANNH if Race recode = American Indian/Alaskan Native and Origin
recode NHIA = Non-Spanish-Hispanic-Latino

• Raceeth = APINH if Race recode = Asian or Pacific Islander and Origin recode NHIA =
Non- Spani sh-Hi spanic-Latino

• Raceeth = Hispanic if Origin recode NHIA = Spanish-Hispanic-Latino

In the Table tab, from the "Available Variables" panel, expand "Merged" and select "Raceeth."
Click the Column button. These options tabulate the rates by sex and race/ethnicity.

Output tab. For "Display Rates as Cases Per" select "1,000,000." For "Number of Decimal
Places for" select "0.001."

Click the Execute button to run the analyses and create the output Rate Matrix.

Use the Matrix and Export menus to output the race/ethnicity/sex data to a text file.

Incidence by Age

Assuming you are in the same session as above, modify the selections as follows:

Table tab. In the Table tab, from the "Display Variables" panel, select the Row variable "Sex"
and click Remove. Also select the Column variable "Raceeth" and click Remove. Use the menus
"File" and "Dictionary" to create a User-Defined variable AgesH4 that takes the values 0-4
years, 5-9 years, 10-14 years, 15-19 years, and 0-19 years if the "Age at Diagnosis" is in the
selected age group. Age at Diagnosis is defined by the values of "Age recode with <1 year olds."
In the Table tab, from the "Available Variables" panel, expand "User-Defined" and select
"AgesH4." Click the Row button. This tabulates the rates by the selected age groups.

DRAFT America's Children and the Environment
December 2021

Page 7

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Click the Execute button to ran the analyses and create the output Rate Matrix.

Use the Matrix and Export menus to output the data to a text file. Tabulate the rates for selected
age groups.

Mortality by Race/Ethnicity and Sex
Click the Rate Session button "E".

Data tab. Select "Mortality - All COD, Aggregated With State, Total U.S. (1990-2018)
-------
Health: Childhood Cancer

In the Table tab, from the "Available Variables" panel, expand "Merged" and select "Raceeth2"
Click the Column button. These options tabulate the rates by sex and race/ethnicity.

Output tab. For "Display Rates as Cases Per" select "1,000,000." For "Number of Decimal
Places for" select "0.001."

Click the Execute button to run the analyses and create the output Rate Matrix.

Use the Matrix and Export menus to output the race/ethnicity/sex data to a text file.

Questions and Comments

Questions regarding these methods, and suggestions to improve the description of the methods,
are welcome. Please use the "Contact Us" link at the bottom of any page in the America's
Children and the Environment website.

Statistical Comparisons

Statistical analyses of the cancer incidence and mortality rates were used to determine whether
the differences between rates for different demographic groups were statistically significant. For
these analyses, the incidence and mortality counts and populations for each combination of
calendar year, age group, sex, and race/ethnicity were obtained from the SEER database using
the same race/ethnicity groups described above (excluding the All category)1, and the age groups
0-4 years, 5-9 years, 10-14 years, and 15-19 years. Incidence data for malignant cancers for the
years 1992-2018 were obtained from the SEER data set "Incidence - SEER Research Data, 13
Registries, Nov 2020 Sub (1992-2018)." U.S. mortality data for 1992-2018 were obtained from
the SEER data set "Mortality - All COD, Aggregated With State, Total U.S. (1990-2018)
-------
Health: Childhood Cancer

population of that demographic subgroup multiplied by the rate, and a variance equal to the mean
multiplied by a dispersion parameter. The logarithm of the rate was assumed to be the sum of
explanatory terms for age, sex, and race/ethnicity. Using this model, the difference in the value
of a rate between different demographic groups is statistically significant if the difference
between the corresponding sums of explanatory terms is statistically significantly different from
zero. A p-value at or below 0.05 implies that the difference is statistically significant at the 5%
significance level. No adjustment is made for multiple comparisons.

For each type of comparison, we present unadjusted and adjusted analyses. The unadjusted
analyses compare a rate between different demographic groups. The adjusted analyses add other
demographic explanatory variables to the statistical model and use the statistical model to
account for the possible confounding effects of these other demographic variables. For example,
the unadjusted race/ethnicity comparisons compare the rates between different race/ethnicity
pairs. The adjusted analyses add age and sex terms to the statistical model and compare the rates
between different race/ethnicity pairs after accounting for the effects of the other demographic
variables. For example, if Hispanic children tend to be younger than White non-Hispanics, and if
the rate strongly depends on age only, then the unadjusted differences between these two
race/ethnicity groups would be significant but the adjusted difference (taking into account age)
would not be significant.

Comparisons of incidence rates between pairs of race/ethnicity groups are shown in Table 1.
Comparisons of mortality rates between pairs of race/ethnicity groups are shown in Table 2. In
Tables 1 and 2, for the "Unadjusted" comparisons, the only explanatory variables are terms for
each race/ethnicity group. For these unadjusted comparisons, the statistical tests compare the
rates for each pair of race/ethnicity groups. For the "Adjusted for age, sex" comparisons, the
explanatory variables are terms for each race/ethnicity together with terms for each age group
and sex. For these adjusted comparisons, the statistical test compares the pair of race/ethnicity
groups after accounting for any differences in the age and sex distributions between the
race/ethnicity groups.

Additional comparisons are shown in Table 3 for incidence rates and in Table 4 for mortality
rates. The Against = "age" unadjusted p-value compares the rates between all the age groups.
The adjusted p-value includes adjustment terms for sex and race/ethnicity in the model. The
Against = "sex" unadjusted p-value compares the rates between the two sexes. The adjusted p-
value includes adjustment terms for age and race/ethnicity in the model. The Against = "year"
unadjusted p-value compares the trends in the rates by regressing against the calendar year. The
adjusted p-value includes adjustment terms for age, sex, and race/ethnicity in the model. To
evaluate whether the difference in rates for Males and Females varies according to the
race/ethnicity, p-values for comparing the Male and Female rates by race/ethnicity group are
found in the rows where Against = "sex" and Subset = the race/ethnicity group. The adjusted p-
value includes an adjustment term for age.

For more details on these statistical analyses, see the memorandum by Cohen (2011).111

111 Cohen, J. 2011. Selected statistical methods for testing for trends and comparing years or demographic groups in
other ACE health-based indicators. Memorandum submitted to Dan Axelrad, EPA, 16 June, 2011.

DRAFT America's Children and the Environment
December 2021

Page 10

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Table 1. Statistical significance tests comparing cancer incidence rates for children ages 0 to 19
years, between pairs of race/ethnicity groups, for 2016-2018.

P-VALUES

Variable

First race/ethnicity
group

Second race/ethnicity group

Unadjusted

Adjusted for age, sex

Cancer incidence

White non-Hispanic

Black non-Hispanic

0.002

<0.001

Cancer incidence

White non-Hispanic

API non-Hispanic

0.114

0.001

Cancer incidence

White non-Hispanic

AIAN non-Hispanic

0.561

0.241

Cancer incidence

White non-Hispanic

Hispanic

0.092

0.001

Cancer incidence

Black non-Hispanic

API non-Hispanic

0.227

0.011

Cancer incidence

Black non-Hispanic

AIAN non-Hispanic

0.571

0.219

Cancer incidence

Black non-Hispanic

Hispanic

0.076

<0.001

Cancer incidence

API non-Hispanic

AIAN non-Hispanic

0.990

0.951

Cancer incidence

API non-Hispanic

Hispanic

0.728

0.380

Cancer incidence

AIAN non-Hispanic

Hispanic

0.902

0.780

Table 2. Statistical significance tests comparing cancer mortality rates for children ages 0 to 19
years, between pairs of race/ethnicity groups, for 2016-2018.

P-VALUES

Variable

First race/ethnicity
group

Second race/ethnicity group

Unadjusted

Adjusted for age, sex

Cancer mortality

White non-Hispanic

Black non-Hispanic

0.001

<0.001

Cancer mortality

White non-Hispanic

API non-Hispanic

0.509

<0.001

Cancer mortality

White non-Hispanic

AIAN non-Hispanic

0.454

<0.001

Cancer mortality

White non-Hispanic

Hispanic

0.761

0.041

Cancer mortality

Black non-Hispanic

API non-Hispanic

0.016

<0.001

Cancer mortality

Black non-Hispanic

AIAN non-Hispanic

0.847

0.316

Cancer mortality

Black non-Hispanic

Hispanic

0.002

<0.001

Cancer mortality

API non-Hispanic

AIAN non-Hispanic

0.326

<0.001

Cancer mortality

API non-Hispanic

Hispanic

0.637

0.012

Cancer mortality

AIAN non-Hispanic

Hispanic

0.413

<0.001

Table 3. Other statistical significance tests comparing cancer incidence rates for children ages 0
to 19 years for 2016 to 2018 (trends for 1992-2018).

P-VALUES

Variable

From

Against

Subset

Unadjusted

Adjusted*

Cancer incidence

2016

2018

age

<0.001

Cancer incidence

2016

2018

race

0.036

<0.001

Cancer incidence

2016

2018

sex

0.038

<0.001

Cancer incidence

2016

2018

sex

White non-Hispanic

0.431

0.003

Cancer incidence

2016

2018

sex

Black non-Hispanic

0.125

0.010

Cancer incidence

2016

2018

sex

API non-Hispanic

0.779

0.550

Cancer incidence

2016

2018

sex

AIAN non-Hispanic

0.556

0.536

DRAFT America's Children and the Environment
December 2021

Page 11

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

P-VALUES

Variable

From

Against

Subset

Unadjusted

Adjusted*

Cancer incidence

2016

2018

sex

Hispanic

0.163

<0.001

Cancer incidence

1992

2018

year

<0.001

*For Against = "age," the p-values are adjusted for sex and race/ethnicity.
For Against = "sex," the p-values are adjusted for age and race/ethnicity.
For Against = "year," the p-values are adjusted for age, sex, and race/ethnicity.

Table 4. Other statistical significance tests comparing cancer mortality rates for children ages 0
to 19 years for 2016 to 2018 (trends for 1992-2018).

P-VALUES

Variable

From

Against

Subset

Unadjusted

Adjusted*

Cancer mortality

2016

2018

age

<0.001

Cancer mortality

2016

2018

race

0.004

<0.001

Cancer mortality

2016

2018

sex

0.042

<0.001

Cancer mortality

2016

2018

sex

White non-Hispanic

0.325

<0.001

Cancer mortality

2016

2018

sex

Black non-Hispanic

0.351

<0.001

Cancer mortality

2016

2018

sex

API non-Hispanic

0.441

<0.001

Cancer mortality

2016

2018

sex

AIAN non-Hispanic

0.304

<0.001

Cancer mortality

2016

2018

sex

Hispanic

0.390

<0.001

Cancer mortality

1992

2018

year

<0.001

Data Files

The following files are needed to calculate this indicator. All these files together with the
SEER* Stat software used to calculate the incidence and mortality rates are available from the
SEER website: http://seer.cancer.gov/data/access.html. The data and analyses used for this
indicator were obtained using version 8.3.9.2 of SEER*Stat.

DRAFT America's Children and the Environment
December 2021

Page 12

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Data Summary

Data used to support Indicators:

H5. Cancer incidence for children ages 0 to 19 years, by type, 1993-2018.
H4. Cancer incidence and mortality for children ages 0 to 19 years, 1992-2018.

Data

Cancer incidence of malignant cancers in SEER 13 registries for children ages 0 to 19
years.

Years

1993

1994

1995

1996

1997

1998

1999

Cancer
incidence

1,723

1,654

1,684

1,775

1,714

1,833

1,779

Years

2000

2001

2002

2003

2004

2005

2006

Cancer
incidence

1,838

1,912

1,982

1,807

1,934

2,029

1,837

Years

2007

2008

2009

2010

2011

2012

2013

Cancer
incidence

2,007

2,036

2,103

2,116

2,093

2,061

2,043

Years

2014

2015

2016

2017

2018

Cancer
incidence

2,159

2,229

2,143

2,054

2,047

DRAFT America's Children and the Environment
December 2021

Page 14

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Calculation of Indicator

All the calculations for this indicator were carried out using SEER*Stat software, version 8.3.9.2.
In this section we detail the menu options required for carrying out these analyses.

Note that any User-Defined variable previously created using the File and Dictionary menus in
the same or any earlier session will not need to be recreated.

ICCC site recode ICD-Q-3/WHO 2008.

Click the Rate Session button "E".

Data tab. Select "Incidence - SEER Research Data, 13 Registries, Nov 2020 Sub (1992-2018)."

Table tab. Use the menus "File" and "Dictionary" to create a User-Defined variable YearsH5
that takes the values 1993-1994, 1995-1996, 1997-1998, 1999-2000, 2001-2002, 2003-2004,
2005-2006, 2007-2008, 2009-2010, 2011-2012, 2013-2014, 2015-2016, and 2017-2018 if the
"Year of Diagnosis" is in the selected year group. In the Table tab, from the "Available
Variables" panel, expand "User-Defined" and select "YearsH5." Click the Row button. From the
"Available Variables" panel, expand "Site and Morphology" and select "ICCC site recode ICD-
O-3/WHO 2008." Click the Column button. This tabulates the rates by the selected calendar year
groups and cancer types.

Output tab. For "Display Rates as Cases Per" select "1,000,000." For "Number of Decimal
Places for" select "0.001."

Click the Execute button to run the analyses and create the output Rate Matrix.

Use the Matrix and Export menus to output the data to a text file. Tabulate the rates for selected
cancer types and year groups.

AYA Site Recode/WHO 2008.

The cancer types acute lymphocytic leukemia and Wilms' tumor are not specifically included in
the ICCC site recode ICD-O-3/WHO 2008 categories. Thus the incidence rate for these two

DRAFT America's Children and the Environment
December 2021

Page 15

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

cancers is calculated separately using the "AYA site recode/WHO 2008," which codes cancer
types for adolescents and young adults.

Modify the above selections as follows:

Table tab. In the Table tab, from the "Display Variables" panel, select the Column variable
"ICCC site recode ICD-0-3/WH0 2008" and click Remove. From the "Available Variables"
panel, expand "Site and Morphology" and select "AYA site recode/WHO 2008." Click the
Column button. This tabulates the rates by the selected calendar year groups for the cancer types
coded in the AYA site recode.

Click the Execute button to run the analyses and create the output Rate Matrix.

Use the Matrix and Export menus to output the AYA site recode data to a text file. Tabulate the
rates for selected cancer types and year groups.

Age-adiustment

For indicator H5, all cancer incidence rates were age-adjusted to the 2000 U.S. Standard
Population using five age groups for children. The calculations were carried out automatically by
the SEER* Stat software. The age-adjusted rates are often preferred when comparing different
populations (such as for different calendar years or different race groups) since differences in the
age-adjusted cancer rates are mainly attributable to factors other than age. In general, a younger
population will tend to have a lower crude (unadjusted) cancer rate. The age-adjustment replaces
the distribution of ages in each given population by the same standard age distribution, in this
case the age distribution of the U.S. population in 2000.

The calculation is shown by the following example which gives the age-adjusted cancer
incidence rate for 2017-2018 for acute myeloid leukemia in children ages 0 to 19 years.

Age-adjusted Acute Myeloid Leukemia Incidence for 2017-2018.

Age

Cases

SEER 13

Crude Rate

U.S. 2000

Adjusted

Group

12]

Population [3]

(Cases per

Standard

Proportion

Rate

11]

Million) [4]

Population

of Age

(Cases

= [2]/[3]

[5]

Group
0-19 [6]

per
Million)
= [4]x
[6]

00 years

1,044,300

22.024

3,794,901

0.048

1.061

01-04

years

4,319,801

14.121

15,191,619

0.193

2.723

05-09

years

5,464,387

5.307

19,919,840

0.253

1.342

10-14

years

5,573,113

6.639

20,056,779

0.255

1.690

DRAFT America's Children and the Environment
December 2021

Page 16

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Age

Cases

SEER 13

Crude Rate

U.S. 2000

Adjusted

Group

12]

Population [3]

(Cases per

Standard

Proportion

Rate

11]

Million) [4]

Population

of Age

(Cases

= [2]/[3]

[5]

Group

per

0-19 [6]

Million)

= [4]x

[6]

15-19

years

5,562,810

10.067

19,819,518

0.252

2.533

00-19

years

206

21,964,411

9.379

78,782,657

1.000

9.348

The first three columns give the number of new malignant acute myeloid leukemia cases in
2017-2018 and the total population for each children's age group within the counties included in
the SEER 13 cancer registries. The number of cases divided by the SEER 13 population and
multiplied by 1 million gives the crude rate shown in the fourth column. For example, for the age
group 00 years, there were 23 cases in a population of 1,044.300 giving a crude rate of
(23/1044300) x 1000000 = 22.024 cases per million. The crude rate for all children ages 0 to 19
years is 9.379 (bottom row).

The U.S. 2000 Standard Population for children is shown in the fifth column. This is the age
distribution for the entire United States in 2000. For analyzing children's populations only five
age groups are needed (0, 1-4, 5-9, 10-14, and 15-19). However, for analyzing children and
adults the U.S. 2000 Standard Population has 19 age groups, since the U.S. 2000 Standard
Population also includes five-year age groups up to 80-84 and the 85+ age group. The standard
age distribution for children 0 to 19 years is shown in the sixth column. The U.S. 2000 Standard
Population in each age group is divided by the total U.S. 2000 Standard Population for children
ages 0 to 19 years, i.e., 78,782,657.

For this example, the crude rate for ages 0 to 19 years, 9.379, is very close to the age-adjusted
rate, 9.348, reflecting the fact that the age distribution for children 0 to 19 years in the SEER 13
regions closely matches the 2000 U.S. population age distribution. For other years or for
different cancer types, crude and age-adjusted rates can differ by a much greater percentage.

Relative Standard Error

Relative Error (%) = [Standard Error (Rate) / Rate] x 100%

DRAFT America's Children and the Environment
December 2021

Page 17

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Rates with a relative error less than 30% were treated as being reliable and were tabulated. For
indicator H5, all the relative standard errors in Table H5 were at most 30% and so all values
were tabulated. For Table H5a (see below), some relative standard errors were greater than 30%
and less than 40% and those rates were tabulated but flagged as to be interpreted with caution.
Rates with a relative standard error greater than or equal to 40%, or without an estimated relative
standard error, were treated as being unreliable; these values were not tabulated and were flagged
as having a large uncertainty.

Rates by Age Group

For Table H5a, cancer incidence for 2017-2018 is tabulated by cancer type and age group.
The Table H5a was calculated as follows using SEER*Stat:

ICCC site recode ICD-Q-3/WHO 2008.

Click the Rate Session button "E".

Data tab. Select "Incidence - SEER Research Data, 13 Registries, Nov 2020 Sub (1992-2018)."

Selection tab. Check "Malignant Behavior." For "Age at Diagnosis" click the Edit button and
then select the age group 0 to 19 years using "Age recode with <1 year olds" = 00 years, 01-04
years, 05-09 years, 10-14 years, and 15-19 years. For "Race, Sex, Year Dx" click the Edit button
and select the years 2017-2018 using "Year of Diagnosis" = 2017 and 2018. This restricts the
data to malignant cancers in children ages 0 to 19 years diagnosed in 2017-2018.

Table tab. Use the menus "File" and "Dictionary" to create a User-Defined variable AgesH5 that
takes the values 0-4 years, 5-9 years, 10-14 years, 15-19 years, and 0-19 years if the "Age at
Diagnosis" is in the selected age group. Age at Diagnosis is defined by the values of "Age recode
with <1 year olds." In the Table tab, from the "Available Variables" panel, expand "User-
Defined" and select "AgesH5." Click the Row button. From the "Available Variables" panel,
expand "Site and Morphology" and select "ICCC site recode ICD-O-3/WHO 2008." Click the
Column button. This tabulates the rates by the selected age groups and cancer types.

Output tab. For "Display Rates as Cases Per" select "1,000,000." For "Number of Decimal
Places for" select "0.001."

Click the Execute button to run the analyses and create the output Rate Matrix.

Use the Matrix and Export menus to output the data to a text file. Tabulate the rates for selected
cancer types and age groups.

DRAFT America's Children and the Environment
December 2021

Page 18

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

AYA Site Recode/WHO 2008.

The cancer types acute lymphocytic leukemia and Wilms' tumor are not specifically included in
the ICCC site recode ICD-0-3/WH0 2008 categories. Thus the incidence rate for these two
cancers is calculated separately using the "AYA site recode/WHO 2008," which codes cancer
types for adolescents and young adults.

Modify the above selections as follows:

Table tab. In the Table tab, from the "Display Variables" panel, select the Column variable
"ICCC site recode ICD-O-3/WHO 2008" and click Remove. From the "Available Variables"
panel, expand "Site and Morphology" and select "AYA site recode/WHO 2008." Click the
Column button. This tabulates the rates by the selected age groups for the cancer types coded in
the AYA site recode/WHO 2008.

Click the Execute button to run the analyses and create the output Rate Matrix.

Use the Matrix and Export menus to output the AYA site recode data to a text file. Tabulate the
rates for selected cancer types and age groups.

Questions and Comments

Statistical Comparisons

Statistical analyses of the cancer incidence rates for each cancer type were used to determine
whether the differences between rates for different demographic groups were statistically
significant. Each cancer type was analyzed separately. Incidence data for malignant cancers in
1993-2018 were obtained from the SEER data set "SEER Research Data, 13 Registries, Nov
2020 Sub (1992-2018)." For these analyses, the incidence counts and populations for each
combination of cancer type, calendar year, age group, sex, and race/ethnicity were obtained from
the SEER database using the age groups 0-4 years, 5-9 years, 10-14 years, and 15-19 years and
the following race/ethnicity groups defined by the Race recode and Origin recode NHIA
variables in the SEER database:

• Raceeth = WhiteNH if Race recode = White and Origin recode NHIA = Non-Spanish-
Hispanic-Latino

• Raceeth = BlackNH if Race recode = Black and Origin recode NHIA = Non-Spanish-
Hispanic-Latino

• Raceeth = AIANNH if Race recode = American Indian/Alaskan Native and Origin
recode NHIA = Non-Spanish-Hispanic-Latino

DRAFT America's Children and the Environment
December 2021

Page 19

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

• Raceeth = APINH if Race recode = Asian or Pacific Islander and Origin recode NHIA =
Non- Spani sh-Hi spanic-Latino

• Raceeth = Hispanic if Origin recode NHIA = Spanish-Hispanic-Latino

Using a Poisson regression model adjusted for dispersion, the number of cases for each
demographic subgroup was assumed to have a Poisson distribution with a mean equal to the
population of that demographic subgroup multiplied by the rate, and a variance equal to the mean
multiplied by a dispersion parameter. The logarithm of the rate was assumed to be the sum of
explanatory terms for age, sex, and race/ethnicitylv. Using this model, the difference in the value
of a rate between different demographic groups is statistically significant if the difference
between the corresponding sums of explanatory terms is statistically significantly different from
zero. A p-value at or below 0.05 implies that the difference is statistically significant at the 5%
significance level. No adjustment is made for multiple comparisons.

For each type of comparison, we present unadjusted and adjusted analyses. The unadjusted
analyses compare a rate between different demographic groups. The adjusted analyses add other
demographic explanatory variables to the statistical model and use the statistical model to
account for the possible confounding effects of these other demographic variables. For example,
the unadjusted age comparisons compare the rates between the different age groups. The
adjusted analyses add sex and race/ethnicity terms to the statistical model and compare the rates
between different age groups after accounting for the effects of the other demographic variables.
For example, if Hispanic children tend to be younger than White non-Hispanics, and if the
cancer incidence rate strongly depends on race/ethnicity only, then the unadjusted differences
between age groups would be significant because of their different race/ethnicity breakdowns but
the adjusted difference (taking into account race/ethnicity) would not be significant.

The statistical comparisons for each cancer type are shown in Table 1. The Against = "age"
unadjusted p-value compares the rates between all the age groups for 2017-2018. The adjusted p-
value includes adjustment terms for sex and race/ethnicity in the model. The Against = "year"
unadjusted p-value compares the trends in the rates for 1993 -2018 by regressing against the
calendar year. Instead of using the two-year rates given in the tables and graphs, the statistical
trend analysis used the annual rates to improve the precision of the trend estimates and p-values.
The adjusted p-value includes adjustment terms for age, sex, and race/ethnicity in the model.

For more details on these statistical analyses, see the memorandum by Cohen (201 l).v

lv For the incidence data used for these analyses, the race recode variable includes an "Unknown" category. For that
category, the SEER database provides case counts, but the reported populations for that category are zero and the
corresponding age-adjusted incidence rates could not be calculated. For this reason, the statistical analysis excludes
data from persons of unknown race.

v Cohen, J. 2011. Selected statistical methods for testing for trends and comparing years or demographic groups in
other ACE health-based indicators. Memorandum submitted to Dan Axelrad, EPA, 16 June, 2011.

DRAFT America's Children and the Environment
December 2021

Page 20

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Table 1. Statistical significance tests comparing cancer incidence rates for children ages 0 to 19
between age groups for 2017 to 2018 and analyzing trends for 1993-2018.

P-VALUES

Type of Cancer

From

Against

Unadjusted

Adjusted*

Acute lymphoblastic leukemia

2017

2018

age

<0.001

Acute lymphoblastic leukemia

1993

2018

year

0.569

0.043

Central nervous system tumors

2017

2018

age

0.931

0.930

Central nervous system tumors

1993

2018

year

<0.001

Soft tissue sarcomas

2017

2018

age

0.208

NA**

Soft tissue sarcomas

1993

2018

year

0.693

0.320

Hodgkin's lymphoma

2017

2018

age

<0.001

Hodgkin's lymphoma

1993

2018

year

0.288

0.066

Acute myeloid leukemia

2017

2018

age

0.001

<0.001

Acute myeloid leukemia

1993

2018

year

0.071

0.083

Non-Hodgkin's lymphoma

2017

2018

age

0.132

0.067

Non-Hodgkin's lymphoma

1993

2018

year

0.293

0.021

Neuroblastoma

2017

2018

age

<0.001

Neuroblastoma

1993

2018

year

0.009

0.011

Thyroid carcinoma

2017

2018

age

NA**

0.007

Thyroid carcinoma

1993

2018

year

0.932

0.704

Osteosarcoma

2017

2018

age

<0.001

Osteosarcoma

1993

2018

year

0.485

0.619

Wilms' tumor

2017

2018

age

<0.001

Wilms' tumor

1993

2018

year

0.913

0.349

Other and unspecified carcinomasf

2017

2018

age

1.000

NA**

Other and unspecified carcinomasf

1993

2018

year

<0.001

NA**

Ewing's sarcoma

2017

2018

age

0.343

0.301

Ewing's sarcoma

1993

2018

year

<0.001

*For Against = "age," the p-values are adjusted for sex and race/ethnicity.

For Against = "year," the p-values are adjusted for age, sex, and race/ethnicity.

**NA. Statistical model failed to converge.

|"MOther and unspecified carcinomas" represents all carcinomas and other malignant epithelial neoplasms other than thyroid carcinoma and
malignant melanoma.

Data Files

The following files are needed to calculate this indicator. All these files together with the
SEER*Stat software used to calculate the incidence rates are available from the SEER website:
http://seer.cancer.gov/data/access.html. The data and analyses used for this indicator were
obtained using version 8.3.9.2 of SEER*Stat.

DRAFT America's Children and the Environment
December 2021

Page 21

DO NOT QUOTE OR CITE

-------
Health: Childhood Cancer

Utah; Atlanta, Georgia; Detroit, Michigan; San Francisco-Oakland, California; Seattle-
Puget Sound, Washington; Los Angeles, California; San Jose-Monterey, California;
Rural Georgia; and the Alaska Native Tumor Registry. Downloaded December 2021.

The databases incorporated into the SEER* Stat software include census population data used to
compute the incidence rates and to age-adjust the rates to the 2000 U.S. Standard Population.
These population data are obtained by NCI from the Census Bureau. Age-adjustment is
explained above.

DRAFT America's Children and the Environment
December 2021

Page 22

DO NOT QUOTE OR CITE

-------