22
PROXIMITY TO
HAZARDOUS WASTE SITES AND
B-CELL LYMPHOPROLIFERATIVE DISEASES:
A PROTOCOL FOR POPULATION-BASED
CASE-CONTROL ANALYSIS OF
EXISTING DATA
DC MIDDLETON, MH WARD, BA SLADE, CV LEE
AND MD LEWIN
This chapter is adapted from a protocol developed
jointly by the Agency for Toxic Substances and Disease Registry (ATSDR)
and the National Cancer Institute (NCI) in response to a recommendation
from the workshop. The protocol is being submitted for IRB approval.
INTRODUCTION
To assess the potential risk for a B-cell neoplasm
associated with exposure to hazardous waste, the workshop epidemiology
working group proposed a case/control study that would combine ATSDR's
information on hazardous waste sites with health data from the NCI. Neoplasms
that predominantly originate from B-lymphocytes include chronic lymphocytic
leukemia (CLL), non-Hodgkin’s lymphoma (NHL), and multiple myeloma (MM).
The following specific question will be addressed in this case-control
analysis: “Is there an association between residential proximity to hazardous
waste sites and the risk of CLL, NHL, or MM?”
BACKGROUND INFORMATION
The Resource Conservation and Recovery Act (RCRA)
In 1976, Congress enacted RCRA, the first comprehensive federal attempt
to deal with the problem of solid waste. A waste is considered to be hazardous
if it has the following characteristics: ignitability, corrosivity, reactivity,
or toxic extraction procedures.
Wastes that are exempted include: household wastes;
agricultural wastes that are returned to the ground as fertilizer; mining
overburden that is returned to the mine site; utility wastes from coal
combustion; oil and natural gas explosion drilling waste; and, wastes from
the extraction, benefaction, and processing of ores and minerals (eg, coal,
cement kiln dust wastes, arsenic
treated wastes, and certain chromium bearing wastes).
In 1984, the act was amended to apply the statute
to “small quantity” producers. RCRA does not address abandoned or
inactive sites. It does not regulate radioactive wastes, which are defined
and covered by the Atomic Energy Act of 1954. RCRA applies to generators
of hazardous wastes, waste transporters, and treatment/storage/disposal
facilities.
The Comprehensive Environmental Response, Compensation,
and Liability Act (CERCLA, Superfund)
In December 1980, Congress enacted the first comprehensive
federal law to respond to releases of hazardous substances to the environment.
The goal of Superfund is to provide immediate cleanup of hazardous waste
contamination from an accidental spill or from chronic environmental damage
(eg, an abandoned hazardous waste disposal site).
Under Superfund, the EPA revised and expanded the
National Contingency Plan (NCP). The revised plan is formally known as
the National Oil and Hazardous Substances Pollution Contingency Plan. The
NCP lists several phases for response to releases of hazardous substances,
pollutants, and contaminants. Phase I is the discovery and notification
phase: sites identified here make up the CERCLA Information System (CERCLIS)
listing.
Superfund also required EPA to establish a National
Priorities List (NPL). Sites added to the NPL are considered priority
sites for remedial action.
Toxic Release Inventory (TRI)
The TRI was authorized by Title III (Emergency Planning
and Community Right-to-Know Act) of the Superfund Amendments and Reauthorization
Act (SARA) of 1986. This act regulates facilities with 10 or more full-time
employees that release more than threshold quantities of covered chemicals.
The TRI contains information on environmental releases (to air, water,
land, or underground injection) for approximately 300 chemicals (or chemical
categories) from facilities in Standard Industrial Classification (SIC)
Codes 20 through 39. EPA is required to collect these data nationally on
an annual basis.
METHODS
Rationale for Study Design
In environmental epidemiology it is often difficult
and expensive to assemble “exposed” and “unexposed” cohorts large enough
to compare the risk of a specific outcome. For these rare outcomes, there
is considerably more efficiency in a population-based case-control study
that assembles cases from registries (or other methods of “complete” case
ascertainment) and compares them to controls randomly selected from the
same general population. The case-control is especially appropriate
for the study described in this protocol, since health data is available
from the NCI and exposure data from ATSDR.
Exposures (including potential confounders) were
assessed retrospectively by direct interview (in person or by telephone)
with the participant (case or control) or a proxy if the subject is deceased
or too ill to participate. A surrogate measure of exposure to hazardous
waste sites (site distance) will be determined after mapping residential
addresses and hazardous waste sites using the geographic information system
(GIS) system. While site distance is an imperfect surrogate measure of
exposure, it is considered both economical and relatively unbiased (1).
Case Definition (Dependent Variables of Interest)
The outcomes of interest are CLL, NHL, and multiple
myeloma (MM). The three variables representing these outcomes
will be dichotomous (dependent) variables for disease status. In essence,
there will be a separate analysis for each of the three outcomes (dependent
variables) and a combined analysis to assess the association between exposure
and developing any B-cell neoplasm (CLL, NHL, or MM).
Exposure Classification (Primary Independent Variable)
CERCLIS sites, RCRA sites, NPL sites, TRI sites,
and residential addresses of participants will be converted into geocoded
data (ie, latitude and longitude). This data will be entered into the ATSDR
GIS system to generate a variable for “site distance,” the distance in
miles between the residences of participants (cases and controls) and the
nearest hazardous waste site(s).
For individual analyses, “site distance” will be
based on the distance from a residence to various types of “hazardous waste
sites.” This approach should make it possible to rank the risk associated
with different types of sites (eg, do NPL sites pose more risk for these
health outcomes than a CERCLIS site not listed on the NPL?).
While the potential for exposure can be divided
into multiple categories (eg, none, low, high), the middle group (as defined
by site distance) is more likely to be affected by misclassification. While
some information may be lost with fewer exposure categories, the validity
of the study may be enhanced using a dichotomous classification of exposure
(1). In the present analysis, the exposure status of cases and controls
will be classified as follows:
(1) “exposed” (site distance £1 mile); or,
(2) “unexposed” (site distance >1 mile).
Data Sources
For the analyses described in this protocol, the
health-related data has already been collected by the NCI and four participating
states during the 1980's in three separate studies:
(1) The Kansas Health and Occupation Study;
(2) A Study of Health and Environmental Factors Affecting Rural Men (The
FARM
Study in Iowa/Minnesota); and,
(3) The Nebraska Health Study (conducted in eastern Nebraska).
These three studies (Iowa/Minnesota, Kansas, and
Nebraska) collected data on cases of certain hematologic malignancies (including
NHL, CLL, and MM) and frequency-matched controls. The studies were all
population-based case-control studies with questionnaires directed towards
potential cancer risk factors, including pesticide use, family medical
history, occupation, use of tobacco products, and alcohol use. While the
questionnaires differed somewhat, the information was collected under similar
circumstances and many of the data elements were similar. NCI has produced
a composite database using information from the three studies.
Risk Factors (Potential Confounders)
Age. The incidence of CLL, NHL, and MM increases
with age. The age of participants will be determined from the “date of
birth” to the “date of interview,” or to the “date of death.”
Cases and controls were frequently matched on age
and all analyses will be adjusted for age using the following categories:
(1) 59 years or less;
(2) 60-69 years;
(3) 70-79 years; and,
(4) 80 or more years.
The factors that follow have been associated with
risk or CLL and/or HNL in these study populations and will be evaluated
as potential confounders and effect modifiers.
Education. To attempt to control for effects
of socioeconomic status, education levels of the participants will be categorized
as follows:
(1) less than 12 years;
(2) equal to 12 years (completed high school); and,
(3) greater than 12 years.
Family History of Cancer. These and other
studies have suggested that an individual's risk of hematologic/lymphatic
cancer is influenced by genetic factors. Patients who develop these malignancies
are more likely to have relatives that with hematologic/lymphatic cancer,
other selected cancers, or even all cancers (2, 3, 4, 5). Family history
(regarding cancer) will be categorized as:
(1) no first degree relatives with cancer;
(2) first degree relatives with other cancers (besides lymphatic/hematopoietic);
and,
(3) first degree relatives with lymphatic/hematopoietic cancer.
Occupation. Some studies have suggested that
farmers are at increased risk for non-Hodgkin's lymphoma (6, 7, 8) and
possibly for multiple myeloma (9). Farming will be evaluated with a dichotomous
variable for ever “living or working on a farm.”
Pesticides. Several pesticides have been
associated with NHL. The “use” of several classes of pesticides will be
characterized as follows:
(1) a dichotomous variable for ever using crop insecticides;
(2) a dichotomous variable for ever using herbicides;
(3) a dichotomous variable for ever using fungicides;
(4) a dichotomous variable for ever using organochlorines;
(5) a dichotomous variable for ever using organophosphates; and,
(6) using 2,4-dichlorophenoxyacetic acid.
Proxy Status. When the case or control is
deceased or too ill to participate, researchers often turn to proxy respondents.
There is, however, evidence that the information from proxy respondents
is not always consistent with information from direct interviews (10).
The effect of proxy status will be evaluated using a dichotomous variable.
Tobacco Products. Tobacco contains numerous
carcinogens and has well established associations with many cancers. A
relationship between tobacco use and leukemia among white men was reported
in a population based case-control study in Iowa and Minnesota (11). The
study (during 1981-1984) included 578 cases and 820 controls. Most pertinent
to the issue
of a monoclonal B-lymphocytosis, the odds ratio comparing the odds
of CLL among tobacco users (in any form) to the odds among nonusers was
1.6 (95% CI= 1.1 - 2.3) (never used tobacco in any form).
Therefore, tobacco use will be categorized as a
dichotomous variable (“ever” or “never” used tobacco).
DATA ANALYSIS
The analyses will proceed as follows:
(1) simple descriptive statistics for each variable;
(2) crude odds ratios associating each risk factor with outcomes of interest;
(3) stratified analyses will be used to calculate adjusted odds ratios
(using the
Mantel-Haenzel estimate) if the Breslow-Day Test is consistent with homogeneity
of
odds ratios across strata; and,
(4) the potential for interaction (effect modification) between exposure
to hazardous
waste and selected risk factors will be evaluated.
Multiple logistic regression will be used to obtain
a summary odds ratio (associating exposure with outcome) while controlling
simultaneously for multiple confounders.
DATA PROTECTION AND PRIVACY
As a federal public health agency, ATSDR is bound
by federal law regarding the protection of the confidentiality of study
participants. Under the Privacy Act of 1974 (5 U.S.C. Section 552a(e)),
employees of federal agencies are responsible for protecting data collected
on identifiable persons or organizations where the supplier of the information
has not given the agency consent to make that data public.
These data will be used only for research and statistical
purposes. No information with personal identifiers (name, address, or social
security number) will be released in the form of a report or publication
and none of the study subjects will be contacted. The protocol will follow
provisions to ensure appropriate confidentiality of personal information.
REFERENCES
1. Shalat SL. Can we use distance to study disease? In: Andrews
JS, Askew LO, Bucsela JA,
Hoffman DA, Johnson BL, Xintaras C,
eds. Proceedings of the Fourth National
Environmental Health Conference. San
Antonio, 1990; 33-38.
2. Cartwright RA, McKinney PA, O'Brien C, Richards DG, Roberts
B, Lauder I, Darwin CM,
Bernard SM, Bird CC. Non-Hodgkin's lymphoma:
case control epidemiological study in
Yorkshire. Leukemia Res 1988; 12(1):
81-8.
3. Linet MS, van Natta ML, Brookmeyer R, Khoury MJ, McCaffrey
LD, Humphrey RL, Szklo
M. Familial cancer history and chronic
lymphocytic leukemia: a case-control study. Am J
Epidemiol 1989; 130(40):655-64.
4. Rodovanovic Z, Markovic-Denic L, Jankovic S. Cancer mortality
of family members of
patients with chronic lymphocytic leukemia.
Kluwer Academic Publishers 1994; 211-13.
5. Ward MH, Zahm SH, Weisenburger DD, Gridley G, Cantor KP,
Saal RC, Blair A. Dietary
factors and non-Hodgkin's lymphoma in
Nebraska (United States). Cancer Causes and
Control 1994; 5:422-32.
6. Cantor KP, AG Everett, Gibson R, Burmeister LF, Brown LM,
Schuman L, Dick FR.
Pesticides and other agricultural risk
factors for non-Hodgkin's lymphoma among men in
Iowa and Minnesota. Cancer Res 1992;
52:2447-55.
7. Pearce N, Bethwaite P. Increasing incidence of non-Hodgkin's
lymphoma: occupational and
environmental factors. Cancer Res 1992;
52(suppl): 5496s-5500s.
8. Scherr PA, Hutchison GB, Neiman RS. Non-Hodgkin's lymphoma
and occupational
exposure. Cancer Res 1992; 52(suppl.)5503s-5509s.
9. Brown LM, Burmeister LF, Everett GD, Blair A. Pesticide exposures
and multiple myeloma
in Iowa men. Cancer Causes & Control.
1993; 4(2):153-6.
10. Johnson RA, Mandel JS, Gibson RW, Mandel JH, Bender AP, Gunderson
PD, Renier CM
. Data on prior pesticide use collected from self-and
proxy respondents. Epidemiology 1993;
4:157-64.
11. Brown LM, Gibson RA, Burmeister LF, Schuman LM, Cantor KP, Fraumeni
JF. Smoking
and risk of leukemia. Am J Epidemiol
1992; 135(7):763-8.
Return To Contents