Assessment of the Occurrence of Cancer
East Harris County, Texas
1995–2012
June 19, 2015

Prepared by the
Texas Department of State Health Services

Investigation# 14004

 Table of Contents
Executive Summary ........................................................................................................................ 1
Background ..................................................................................................................................... 2
Methods........................................................................................................................................... 2
Data Sources ............................................................................................................................... 3
Statistical Analysis ...................................................................................................................... 3
Results ............................................................................................................................................. 4
Discussion ....................................................................................................................................... 5
Recommendations and Next Steps.................................................................................................. 6
Additional Information ................................................................................................................... 6

Tables and Figures
Figure 1. .......................................................................................................................................... 7
Table 1. ........................................................................................................................................... 8
Table 2. ........................................................................................................................................... 9

Investigation# 14004

 Executive Summary
Citizen concern prompted the Environmental and Injury Epidemiology and Toxicology Unit
(EIET) and Texas Cancer Registry (TCR) of the Texas Department of State Health Services
(DSHS) to examine the occurrence of cancer in east Harris County, Texas.
DSHS followed the Centers for Disease Control and Prevention (CDC) and Council of State and
Territorial Epidemiologists (CSTE) 2013 guidelines to investigate the occurrence of 17 types of
childhood and adult cancers in a geographic area selected in collaboration with community
members. In accordance with these guidelines, the purpose of this assessment was to determine
whether the observed number of cancer cases is statistically significantly greater than expected.
It was not intended to determine the cause of the observed cancers or identify possible
associations with any risk factors.
DSHS staff analyzed TCR data available for an 18-year period spanning from 1995 to 2012.
United States Census data was used to estimate the population in the selected geographic area,
which consisted of 38 census tracts. To evaluate the occurrence of cancer in the area
investigated, the number of observed cancer cases was compared to what would be expected for
the area based on cancer rates in Texas. Standardized incidence ratios (SIRs) were calculated as
the number of observed cases divided by the number of expected cases in the area of concern for
the 18-year period (1995-2012). A 95 percent confidence interval (CI) was calculated for each
SIR to determine statistical significance.
Observed numbers of several of the 17 cancers analyzed were statistically significantly greater
than expected, while others were statistically significantly less than expected. In accordance with
the CDC and CSTE 2013 guidelines, DSHS will review these results with a group of subject
matter experts to assess the feasibility of follow-up epidemiologic study.

Investigation# 14004

1

 Background
Citizen concern expressed to the Texas Department of State Health Services (DSHS) prompted
the agency’s Environmental and Injury Epidemiology and Toxicology Unit (EIET) and Texas
Cancer Registry (TCR) to examine the occurrence of cancer in east Harris County, Texas. Local
residents were concerned about a possible excess of cancer cases occurring in the flood plain
around the San Jacinto River, which includes the San Jacinto River Waste Pits (SJRWP)
Superfund site.
The Centers for Disease Control and Prevention (CDC) and Council of State and Territorial
Epidemiologists (CSTE) define a cancer cluster as a greater than expected number of cancer
cases that occurs within a group of people in a geographic area over a defined period of time1.
DSHS followed the CDC and CSTE 2013 Guidelines for Investigating Suspected Cancer
Clusters and Responding to Community Concerns1 to investigate the occurrence of cancer in this
community.
The CDC and CSTE guidelines include four steps1. The first step is to collect information about
the community’s concerns. The second step, reported here, is to determine whether the observed
number of cancer cases is statistically significantly greater than expected. It is important to note
that the data and statistical analysis conducted at this step cannot determine if cancers observed
in the community are associated with environmental, lifestyle, or other risk factors.
The guidelines also provide additional steps that can be followed when appropriate. The third
step is to evaluate the feasibility of performing an epidemiologic study to examine if exposure to
a specific risk factor is associated with the suspected cancer cluster, and the fourth step is to
conduct an epidemiologic study, if deemed feasible in step three. Many factors are considered in
making the determination to progress to steps three or four. The CDC and CSTE guidelines state,
“only a small fraction of cancer cluster inquiries might meet the statistical and etiological criteria
to support a cluster investigation through all the steps outlined….”1
Methods
Consistent with the CDC and CSTE guidelines, DSHS collaborated with the community to select
the geographic area, time frame, and cancers to be included in this analysis.
The following cancer types were included in the analysis: childhood leukemia, lymphoma,
kidney (renal), liver (hepatic), brain cancers, glioma (a brain cancer subtype), melanoma, and
retinoblastoma; and all-ages leukemia, lymphoma, myeloma, kidney (renal), liver (hepatic),
brain, breast, cervical, and thyroid cancers.
Complete TCR cancer data are available for 1995 to 2012. DSHS evaluated all 18 years of
available cancer data because the community was concerned about cancers occurring during the
entire time period.

1

Centers for Disease Control and Prevention, Investigating Suspected Cancer Clusters and Responding to
Community Concerns. MMWR, 2013. 62: p. 22.

Investigation# 14004

2

 The geographic area investigated is consistent with the San Jacinto River surge inundation
zones2 identified by the Environmental Protection Agency (EPA)3. It also includes
neighborhoods identified in a 2013 Public Health Assessment conducted by DSHS, which
considered the areas investigated by the EPA4. The 38 census tracts comprising the area
investigated are shown in Figure 1.
This document outlines the results from step two of the CDC and CSTE guidelines, and only
addresses the question, “Is there a statistically significant excess of cancer in the area of
investigation?”
Data Sources
For each cancer type, the number of cases observed from 1995 to 2012 in the area included in the
investigation was obtained from the TCR (Incidence – Texas, 1995-2012, SEER*Prep 2.5.2).
The TCR is responsible for the collection, maintenance, and dissemination of high-quality Texas
population-based cancer data, and meets national CDC timeliness and data quality standards, as
well as North American Association of Central Cancer Registry certification standards.
Childhood cancers (those occurring among individuals ages 0 to 15 years) were defined
according to the International Classification of Childhood Cancer5. Statewide cancer rates for the
same time period were also obtained from the TCR.
Population estimates for 1995 to 2012 were calculated using linear interpolation based on
population counts obtained from the United States Decennial Census6 for the years 1990, 2000,
and 2010. This method, outlined by the United States Census Bureau7, assumed population
growth occurred in a linear manner.
Statistical Analysis
To determine if a statistically significant excess of cancer existed in the area investigated, the
number of observed cancer cases was compared to what would be expected for the area based on
cancer rates in Texas. Characteristics such as race, sex, and age are closely related to cancer. To
ensure that differences between the numbers of observed and expected cancer cases are not
simply due to differences in these demographic characteristics, the expected numbers of cancer
cases were calculated by multiplying the age-, sex-, and race-specific cancer incidence rates of
2

Brody, S, Blessing R, Atoba K, Mobley W, and Wilson M, A Flood Risk Assessment of the San Jacinto River
Waste Pit Superfund Site. 2014, Center for Texas Beaches and Shores: Texas A&M University Galveston.

3

Environmental Protection Agency. San Jacinto River Waste Pits. January 2015; Available from:
http://www.epa.gov/region6/6sf/pdffiles/san-jacinto-tx.pdf.
4

Texas Department of State Health Services, Environmental & Injury and Toxicology Branch, and Agency for
Toxic Substances and Disease Registry, Public Health Assessment, San Jacinto River Waste Pits. 2013, EPA
Facility ID: TXN000606611: Channelview, Harris County, TX.
5

Steliarova-Foucher E, Stiller C, Lacour B, and Kaatsch P, International Classification of Childhood Cancer, third
edition. Cancer, 2005. 103(7): p. 1457-1467.

6

United States Census Bureau. American FactFinder. 2012; Available from:
http://factfinder2.census.gov/faces/nav/jsf/pages/index.xhtml.

7

US Census Bureau. Methodology for the Intercensal Population and Housing Unit Estimates: 2000 to 2010. 2012;
Available from: http://www.census.gov/popest/methodology/2000-2010_Intercensal_Estimates_Methodology.pdf.

Investigation# 14004

3

 Texas residents (reference population) by the number of people in the corresponding
demographic groups in the area of investigation.
Standardized incidence ratios (SIRs) were calculated to determine if an excess of cancer exists in
the area. The SIR is the number of observed cases compared to (divided by) the number of
expected cases for each cancer type. A SIR greater than 1.00 indicates that the observed number
of cases of a specific cancer type is higher than expected and a SIR less than 1.00 indicates that
the observed number of cases of a specific cancer type is lower than expected.
Few, if any, communities will have exactly the same rate as the average state rate for a similar
population; most will be higher or lower. Therefore, 95 percent confidence intervals (CI) were
calculated for the SIRs to determine if the observed number of cases was statistically
significantly different than expected. If a 95 percent CI (range) includes 1.00, no statistically
significant excess (or reduction) of cancer is indicated. If a 95 percent CI does not contain 1.00,
the SIR is outside the expected range and is statistically significant. When using a 95 percent CI,
5 percent of SIR values calculated is expected to be statistically significantly higher or lower
than the state average due to random chance alone.
In all cases, when results are described as significant or not significant, DSHS is referring only to
statistical significance, with the understanding that all cases of cancer are significant to the
individual, the family, and friends of the individuals who are affected.
Results
Table 1 presents the number of observed cases, the number of expected cases, the SIRs, and the
corresponding 95 percent CIs for each cancer type evaluated in the area with all census tracts
analyzed together. The number of childhood lymphoma and melanoma cases observed in the
area investigated was statistically significantly higher than expected. The number of brain and
cervical cancers among all ages was statistically significantly higher than expected. The number
of thyroid cancers for all ages was statistically significantly lower than expected.
SIRs and 95 percent CIs were also calculated for each census tract separately. Table 2 presents
the number of observed cases, the number of expected cases, the SIRs, and the corresponding 95
percent CIs for each individual census tract. For each of these census tracts, only cancers with a
significantly higher or lower than expected number are shown. Non-significant results were not
included in Table 2.





The number of childhood brain cancer, leukemia, melanoma, and glioma cases were
statistically significantly higher than expected in census tracts 2519, 2323, 2330, and 2520,
respectively.
The number of childhood retinoblastoma cases was statistically significantly higher than
expected in census tracts 2328 and 2529.
The number of male breast cancer, kidney cancer, and leukemia cases among all ages were
each statistically significantly higher in 1 of the 38 census tracts.
The number of liver, brain, and cervical cancer cases among all ages was statistically
significantly higher than expected in two, three, and five census tracts, respectively.

Investigation# 14004

4

 


The number of both lymphoma and myeloma cases were statistically significantly higher in
one, and statistically significantly lower in another, of the census tracts.
The number of female breast cancer cases was statistically significantly higher in three, and
statistically significantly lower in eight of the census tracts.

Discussion
Consistent with the second step of the CDC and CSTE guidelines for investigating suspected
cancer clusters, the primary purpose of this step (assessment) is to determine whether the
observed number of cases is statistically significantly greater than expected1. It is not intended to
determine the cause of the observed cancers or identify possible associations with any risk
factors.
The assessment step in a cancer cluster investigation has several inherent limitations, and results
should be interpreted with these limitations in mind. Cancer is not a single disease, but rather
many different diseases. Different types of cancers vary in etiologies (causes or origins) and may
not share the same predisposing factors. Cancers may be associated with a variety of factors such
as genetics, lifestyle, and socioeconomic status. Because cancer is common, cases might appear
to occur with alarming frequencies within a community even when the number of cases is within
the expected rate for the population.
Additionally, cancer incidence data are based on residence at the time of diagnosis. As people
move, it becomes more difficult to determine whether living in the area of investigation is
associated with an excess of cancers, because residential history is not tracked. Latency (the time
period elapsed between exposure and illness onset) adds to the complexity of this step in the
investigation. For most adult cancers, a period of 10 to 40 years can elapse between the
beginning of an exposure to a cancer-causing agent and the development of a clinically
diagnosable case of cancer. It is possible that former residents who developed cancer no longer
lived in the area at the time of diagnosis, and these cases would not be included in this
assessment. It is also possible that new people have moved into the area and then were diagnosed
with cancer; these cases are included in this assessment.
For this assessment, DSHS analyzed cancer incidence for both the 38 census tracts together and
for each of the census tracts separately, as requested by the community. However, the results of
the individual census tract analyses should be interpreted with caution. The numbers of observed
and expected cases for some of the cancer types were small. SIRs based on small numbers often
yield wide confidence intervals, which reduces the reliability of SIR estimates.
Furthermore, the validity of these census tract-level analyses was limited by the statistical issue
of multiple comparisons. Evaluation of 17 different types of cancers for 38 individual census
tracts resulted in 646 census tract-level SIR estimates. Since a 95 percent confidence interval was
used, 5 percent of estimates (32 of the 646 SIRs calculated) may be statistically significant due to
chance alone, not because of a true excess of cancers.
Conclusion
This assessment identified a number of statistically significant results that warrant further
discussion. In particular, the SIRs for childhood glioma in census tract 2520, childhood
melanoma in census tract 2330, and childhood retinoblastoma in census tracts 2328 and 2529 are
Investigation# 14004

5

 notable. Childhood cancers are rare and have shorter latency periods. However, the limitations
mentioned above must be taken into account when interpreting these results. The SIR estimates
for each of these three childhood cancers were based upon very small numbers of cases observed
over a long period of time (18 years). As mentioned above, reliability is reduced for estimates
based on small numbers, even if the SIRs are large. An additional limitation of the childhood
melanoma analysis is the under-reporting of these cases to the TCR. Melanoma diagnoses and
treatment often occurs in an outpatient setting, where automated surveillance systems usually do
not exist. Incomplete data could lead to biased and misleading results.
Recommendations and Next Steps
In keeping with the CDC and CSTE guidelines, DSHS will consult with a group of subject
matter experts to review statistically significant cancers identified in this assessment (especially
the rare childhood cancers with large SIR estimates) and evaluate whether follow-up
epidemiologic study is recommended and feasible. This group will include internal and external
experts in epidemiology, oncology, and toxicology, as well as a citizen to represent the
community’s interests.
Additional Information
For additional information about cancer clusters, visit the Centers for Disease Control and
Prevention, “About Cancer Clusters,” web page at http://www.cdc.gov/nceh/clusters/about.htm.
For additional information on cancer risk factors, visit the American Cancer Society, “What
Causes Cancer?” web page at http://www.cancer.org/cancer/cancercauses/index.
Questions or comments regarding this investigation may be directed to Emily Hall, MPH,
Epidemiologist, Environmental & Injury Epidemiology & Toxicology Unit, at 512-776-3723
(email: emily.hall@dshs.state.tx.us) or to Leticia Nogueira, PhD, MPH, Epidemiology Manager,
Cancer Epidemiology and Surveillance Branch at 512-776-3422 (email:
leticia.nogueira@dshs.state.tx.us).

Investigation# 14004

6

 Figure 1. Selected Census Tracts (2000) for East Harris County.

Note: Some census tracts were subdivided in the 2010 census; these subdivision boundaries are
shown.

Investigation# 14004

7

 Table 1. Standardized Incidence Ratios (SIRs) and 95 percent Confidence Intervals (CIs) for
Selected Cancers in 38 East Harris County Census Tracts Analyzed Together, 1995–2012.
Cancer Type
Childhood
Brain
Glioma
Leukemia
Lymphoma
Melanoma
Hepatic
Renal
Retinoblastoma
All Ages
Brain
Female Breast
Male Breast
Cervix
Kidney
Leukemia
Liver
Lymphoma
Myeloma
Thyroid

Observed‡ Expected‡

SIR

95% CI

55
10
59
44
13
≤5
10
8

52.7
7.7
67.0
30.3
5.4
<1
9.0
4.8

1.04
1.30
0.88
1.45
2.41
1.00
1.11
1.66

(0.79, 1.36)
(0.62, 2.38)
(0.67, 1.14)
(1.05, 1.95)*
(1.28, 4.11)*
(0.21, 2.95)
(0.53, 2.04)
(0.72, 3.27)

232
1917
19
226
491
441
253
651
191
262

200.2
1939.9
13.6
185.6
481.4
413.6
223.8
632.9
186.3
305.8

1.16
0.99
1.39
1.22
1.02
1.07
1.13
1.03
1.03
0.86

(1.01, 1.32)*
(0.94, 1.03)
(0.84, 2.18)
(1.06, 1.39)*
(0.93, 1.11)
(0.97, 1.17)
(1.00, 1.28)
(0.95, 1.11)
(0.89, 1.18)
(0.76, 0.97)†

*Indicates observed number of cancer cases is statistically significantly higher than expected
†Indicates observed number of cancer cases is statistically significantly lower than expected
‡For observed case counts less than five, numbers have been suppressed to protect confidentiality

Investigation# 14004

8

 Table 2. Statistically Significant Standardized Incidence Ratios (SIRs) and 95 percent
Confidence Intervals (CIs) for Selected Cancers in East Harris County by Census Tract, 1995–
2012.
Cancer Type
Childhood
Brain
Leukemia
Glioma
Melanoma
Retinoblastoma

All Ages
Brain

Female Breast

Male Breast
Cervix

Kidney
Leukemia
Liver
Lymphoma
Myeloma

Census Tract Observed‡

Expected‡

SIR

95% CI

2519
2323
2520
2330
2328
2529

7
10
≤5
≤5
≤5
≤5

2.8
4.6
<1.0
<1.0
<1.0
<1.0

2.52
2.17
9.29
8.48
14.35
16.40

(1.01, 5.19)*
(1.04, 3.99)*
(1.12, 33.55)*
(1.02, 30.62)*
(1.74, 51.83)*
(1.99, 59.26)*

2330
2519
2533
2323
2328
2331
2528
2534
2538
2546
2547
3426
2523
2330
2518
2525
2529
2534
2529
2528
2524
2543
2531
2543
2536
2540

18
23
10
72
21
61
77
9
78
23
33
8
≤5
19
7
8
13
≤5
33
21
13
13
≤5
28
≤5
9

9.3
13.5
4.8
91.3
34.8
85.6
60.1
17.9
60.5
35.8
22.0
22.7
<1.0
9.2
1.5
3.3
6.4
<1.0
21.2
11.7
6.8
5.7
<15.0
17.7
<10.0
4.1

1.94
1.71
2.10
0.79
0.60
0.71
1.28
0.50
1.29
0.64
1.50
0.35
5.08
2.06
4.81
2.43
2.02
3.91
1.55
1.80
1.91
2.29
0.43
1.58
0.17
2.20

(1.15, 3.07)*
(1.08, 2.57)*
(1.01, 3.86)*
(0.62, 0.99) †
(0.37, 0.92) †
(0.55, 0.92) †
(1.01, 1.60)*
(0.23, 0.95) †
(1.02, 1.61)*
(0.41, 0.96) †
(1.03, 2.11)*
(0.15, 0.69) †
(1.05, 14.85)*
(1.24, 3.22)*
(1.93, 9.91)*
(1.05, 4.80)*
(1.07, 3.45)*
(1.27, 9.13)*
(1.07, 2.18)*
(1.11, 2.75)*
(1.02, 3.27)*
(1.22, 3.91)*
(0.14, 0.99) †
(1.05, 2.29)*
(0.00, 0.97) †
(1.00, 4.17)*

*Indicates observed number of cancer cases is statistically significantly higher than expected
†Indicates observed number of cancer cases is statistically significantly lower than expected
‡For observed case counts less than five, numbers have been suppressed to protect confidentiality
Investigation# 14004

9

 Investigati0n# 14004

10