Full text of "Evaluation of cancer cluster reports in North Carolina"

See other formats

A SPECIAL REPORT SERIES BY THE N.C. DEPARTMENT OF ENVIRONMENT, HEALTH, AND NATURAL RESOURCES,
DIVISION OF STATISTICS AND INFORMATION SERVICES, CENTER FOR HEALTH AND ENVIRONMENTAL STATISTICS,

P.O. BOX 27687, RALEIGH, N.C. 27611-7687.

No. 56

April 1991

EVALUATION OF CANCER CLUSTER REPORTS
IN NORTH CAROLINA

Tim E. Aldrich, Ph.D.

Julius Lindsey

Peter Morris, M.D.

ABSTRACT

North Carolinians have long been concerned about the patterns of cancer occurrence in the state,
especially those that signal a potential environmental risk. Being increasingly faced with the task of
responding to citizen reports of apparent cancer clusters (defined as the occurrence of an increased rate of
cancer in a small area or within a short period of time), the North Carolina Department of Environment,
Health, and Natural Resources implemented a formal program for the evaluation of the cluster reports in
1982. Mounting volume of cancer cluster reports and limited resources resulted in the development of a
revised protocol involving the Cancer Surveillance Section of the Division of Statistics and Information
Services and use of more databased evaluation procedures.

This report provides an overview of the current procedures in use for evaluating cancer cluster reports
in North Carolina and highlights a few of the most interesting cluster evaluations over the last sixteen
months. Since September 1989, forty-nine reports of cancer clusters have been received from thirty-one
counties. Twenty-seven of these reports have been evaluated and closed. Eleven cluster reports are
currently in the process of active evaluation and another eleven are awaiting the start of evaluation
activities. Of the closed reports, five indicated increased cancer occurrence with three of the five
potentially providing additional clues to environmental risk factors.

INTRODUCTION

A cancer cluster is the occurrence of a greater than
expected number of cases of cancer within a small
geographic area and/or within a short period of time
(i.e., 3-5 years). Cancer clusters are usually reported
when people learn that several friends, family
members, neighbors or co-workers have been
diagnosed with cancer. Because cancer is such a
common disease, it is important to discover whether
the cancers that have been observed truly represent
the occurrence of more cases than one would expect.

The evaluation of cancer cluster reports may help
to determine specific causes of cancer or to identify
risks for cancer in the environment (Caldwell,
1989). Cancer cluster reports also serve an important
social service as well by assuaging public anxiety over
potential environmental cancer risks in their
community (Fiore, et al., 1990). Nearly all state
health departments have developed a protocol for
investigating cancer cluster reports. The procedures
vary greatly, but each mechanism attempts to respond
to very real public concern (Warner and Aldrich,
1987; Thacker, 1989).

With the establishment of the Environmental
Epidemiology Section in October, 1982, North
Carolina began a formal process for evaluating of
cancer cluster reports. This effort was enhanced by
the re-establishment of a statewide cancer registry in
1988.

Classifying Cluster Reports

Cluster reports are divided into three broad
categories: SUSPECTED clusters, REAL clusters
and MEANINGFUL clusters.

SUSPECTED CLUSTERS: When a person reports
a perceived increased cancer occurrence in a small
geographic area, but none is found to exist, then this
report is categorized as a "suspected" cluster. These
reports are not really clusters at all. The observed
cases may appear to represent a cluster partly
because cancer is so common (i.e., it strikes one of
four people sometime during their lifetime).
"Suspected" cluster reports represent about
seventy-five percent of all cancer cluster reports
received.

REAL CLUSTERS: When significantly more
cancers occur than would be expected within a small
geographic area and/or in a short time period, then a
"REAL" cluster is said to exist. However, a REAL

cluster is one for which there is no pattern to the
increase that can suggest an interpretation of potential
increased risk. These REAL clusters will occur
periodically, simply by chance. For Example:

In one of every 256 families of four, all four
members will have cancer sometime during
their lifetime, by chance alone. Assuming a
100-year lifetime, one family of four out of
25,600 families of four could all develop
cancer in a single year. That would be a
REAL cluster, but it would be due to chance.

REAL cluster reports are often found in communities
where there is a large percentage of older persons.
Cancer rates increase with age and adjusting for age
often explains the observation of higher than expected
number of cases. REAL cluster rejxjrts represent
about twenty percent of all reports received.

MEANINGFUL CLUSTERS: When a cluster is
shown to be "REAL" and there is a pattern to the

increase that suggests a group of people at unusually
high risk of cancer, then that cluster is classified as
"MEANINGFUL." These cluster reports comprise
about five percent of all cluster reports received and
are the ones that lead to further epidemiologic
studies.

BACKGROUND

Historically, the North Carolina cancer cluster
evaluation protocol operated from the Environmental
Epidemiology Section of the Department of
Environment, Health, and Natural Resources.
Evaluations of reports of cancer clusters were
generally rigorous scientific activities featuring
extensive field work, often including interviews and
environmental sampling. Some of the routine
procedural steps are noted in the discussion of the
current protocol (see below). All cluster reports
were acknowledged, but usually, only two or three
studies were able to be mounted in a single year. This
protocol was first used in 1 982 when the Environmental
Epidemiology Section conducted a cancer cluster
study in Cherokee and Macon counties. That study
was both time-consuming and incomplete, as only
mortality data were available for the statistical
analysis of the study population. The findings from
the investigation were equivocal.

Following that experience, the Environmental
Epidemiology Section joined with representatives
from universities and the scientific community to
undertake a study of the cancer database needs for

responding to environmental health concerns. With
funding from the Office of Technology Assessment,
this group investigated the cancer databases in other
states and recommended that North Carolina
establish a population-based cancer registry
( Aldrich, et al., 1989). The registry was funded and
is located in the Division of Statistics and Information
Services of the Department of Environment, Health,
and Natural Resources. When the registry became
functional (Fall, 1989), the Environmental
Epidemiology Section began negotiations leading to
a revised protocol for use in the evaluation of cancer
cluster reports (see Figure 1). This protocol was
implemented gradually through the Fall of 1989 and
became fully operational on January 1, 1990. The
experience upon which the present report is based
spans the entire sixteen month period from
September 1, 1989 to December 31, 1990.

MATERIALS AND METHODS

In North Carolina, suspected cancer clusters may
be reported to county health departments, the state
health department, a physician, the Central Cancer
Registry, or the American Cancer Society. Wherever
reports are initially received, they are forwarded to
the Cancer Surveillance Section for evaluation.
These evaluations consist of ten basic steps (NOTE:
An asterisk indicates the step was also a pan of the former
protocol).

Step 1: A letter acknowledging receipt of the
report.* The person reporting a possible
cancer cluster receives written verification
of their report. This letter includes a
description of the steps that will be followed
in the evaluation and an estimate of the
time frame involved. Copies of the letter
are sent to the Environmental Epidemiology
Section and to the appropriate county
health department.

Step 2: Verification of reported cases and a
search for additional cases.* A critical
step in the cluster evaluation process is
identifying all cancer cases in the study
area. Relying solely on the respondent's
report of cases is not adequate. It is extremely
important to verify the accuracy of the
diagnosis for the reported cases and to
determine if any cases are missing. Prior to
1988 a search for cancer cases through
hospital records and local doctors' offices
had to be conducted. In a few communities,
surveys of residents have been performed
to seek additional cases. These instances are

usually those involving small communities
or communities where a large number of
the cancer cases are likely to be still living.
After 1987, cancer cases can be ascertained
from the Central Cancer Registry for many
counties (all North Carolina counties will
be completed in the database for 1990 cases
and subsequent years).

Step 3: Ascertainment of mortality data for the
prior twenty years.* The purpose of this
step is to determine what the trends in
mortality rates have been in and around the
specified area. Some regions of North
Carolina have experienced elevated rates
for specific cancers for many years; generally
these are multiple county areas and represent
an objective of research independent of
cancer cluster reports. The results of these
mortality analyses may influence whether
or not a decision is made to proceed with
the evaluation of a cluster report, e.g., a
cluster report may represent recognition of
a regional trend that is already under study.

This is also a point when adjusting for age
explains many cluster reports received from
areas of the state where there are large
numbers of older, retired citizens. Race and
sex specific age-adjusted mortality rates for
thirty-two cancer sites have been calculated
for all 100 counties of North Carolina for
1970 through 1989. Because many cancers
are quite rare and many county populations
quite small, these rates must be interpreted
cautiously.

Step 4: Crude estimation of cancer incidence. As

cancer survival increases, mortality data
alone becomes less appropriate in deter-
mining whether further evaluation of a
cluster report is justified. Thus, the incidence
of cancer in communities reporting clusters
must often be ascertained. Beginning in
1990, the Central Cancer Registry collects
incidence data on all new cancer cases
statewide, so that in the future estimates of
these rates will no longer be necessary — the
actual rates will be available. For studies of
reported clusters prior to 1990, an estimate
of the expected number of cases is made
(using national rates) to determine what the
expected number of cancer cases would be;
this may then be compared to the number
of cancer cases identified in study area (see
Step 2).

Step 5: Statistical analysis of the incidence and
mortality data. The computer software
CLUSTER was developed in North Carolina
with federal funding from the Agency for
Toxic Substances and Disease Registry
(Aldrich, 1990). This computer program
provides twelve different statistical tests for
analyzing the patterns of occurrence for
rare health events. When the observed
number of cancer cases greatly exceeds the
number of cases expected, this is one
criterion that a cluster exists. Because many
cancers are quite rare, an "increase" may
sometimes only involve two or three cases;
special statistical methods are used for
studying these rare events. Similarly, an
overall cancer rate may not be elevated
(e.g., for a whole county), but one subgroup
of the population may be experiencing
more cancer cases than they should (e.g., a
single neighborhood, black women, etc.),
based on the population in the community.
Analyses by population subgroup are a part
of the CLUSTER capabilities, as well as
studies of spatial patterns and time trends
among cancer cases.

Step 6: Review of the relevant scientific
literature.* As one step for evaluating each
cancer cluster report, scientific articles are
reviewed. The articles studied are related to
the particular cancers that may comprise
the reported cluster. Also, scientific articles
will be reviewed related to health effects
from exposure to an agent or an exposure
source that the community may cite as a
part of the cluster report. Cancer is
comprised of many diverse disease processes
and there are many hypotheses for risk
factors for virtually all cancers. Identifying
known and suspected risk factors of specific
cancers can guide the evaluation of a cluster
report to exposures which might not
otherwise have been suspected. State agency
records are accessible during a cluster
evaluation for determining the location of
specific industries or hazardous waste sites,
particularly those from which there are
potentially carcinogenic emissions.

If cluster report evaluations are to serve
as a means for identifying new environmental
risk factors for cancer, a necessary step is to

assess the presence of risk factors which
may already have been identified from
other cluster studies. Similarly, the results
of MEANINGFUL cluster reports are
published whenever possible so that
researchers in other states can be informed
of what risk factors may have been identified
from North Carolina's cancer cluster
reports.

Step 7: A site visit is always made to the location
reporting a cancer cluster (whether citizens
are contacted or not). These visits are made
in the course of routine travel related to
Central Cancer Registry operations, not as
trips solely for visiting a cluster community.
This approach to site visits is based largely
on financial considerations; as such it can
lead to some cluster sites being visited
sooner than others. On the site visit, a map
is prepared of the entire geographic area,
highlighting the source locations of potential
hazardous exposures. Often, the environ-
mental source identified by the citizen
reporting the cluster is not as "remarkable"
as another one that they may not even be
aware exists.

This is a point in the cluster evaluation
where the literature review may play a
pivotal role; suggested risk factors men-
tioned in the literature may be unknown to
the citizen. This is also when community
education is a consideration; many cancers
have recognized lifestyle, age-specific and
genetic risk factors and may have little
relation to environmental exposures. Such
cancers as lung and breast cancer may
comprise the majority of cases that the
citizen reports in the cluster; yet a few rare
cancers may represent more meaningful
envirorunentally-related events. For these
reasons, all cluster report sites are visited;
most visits are dead ends (hence visits are
made at the convenience of stafO, but
occasionally a previously unsuspected
environmental hazard is identified. That is
also why it is important that all cluster
reports be evaluated; the increased number
of cancer cases that offer a clue to
environmental cancer risks may not be
those that drew the informant's attention
to report the cluster.

Step 8: On-site meeting with area residents (if
requested).* This is the point at which
educational information is provided to
residents about their cluster report and
about cancer in general. These educational
sessions are always coordinated with local
health agencies (e.g., county health
department, American Cancer Society).
This coordination is a effort to promote
broader and continuing cancer educational
activities for local communities.

Step 9: Final report of conclusions and recom-
mendations* These reports are provided
to all parties associated with the cluster
(e.g., county health department); these
reports are "public" documents, available
to all who request a copy. Some reports are
quite detailed, while others are relatively
brief, depending on the availability of data
and the findings of the evaluation. The
information related to personal histories,
like all individual records, are confidential.

Step 10: Entry into a continuing file of Cluster
Reports. All communities that report a
cluster become a site for continuing
surveillance. Based on the statewide database
being created through the Central Cancer
Registry, the cancer experience for each
community will be monitored for a
minimum of five years. The specific cancers
that comprised the cluster report will be
studied, as will other cancers that may have
a relation to environmental exposures.

Close attention is paid when a second
cancer cluster report is received from the
same or a nearby community. The previous
cluster file will be reviewed and the maps
prepared from earlier site visits are inspected.
Occasionally, cluster reports will be
evaluated simultaneously from neighboring
areas, which can provide a cost savings as
well as a more productive process for
evaluating disease patterns.

The duration or timing of the steps of the
evaluation protocol may vary with the individual
' cluster report; however, the sequence is always
maintained (Figure 1). All cancer cluster report
evaluations are monitored through periodic activity
reports, of which this report is one. When a cluster
evaluation results in inconclusive evidence of
clustering, the community represented by the cluster

report will be kept under surveillance for a minimum
of five years. When an evaluation indicates a
potentially MEANINGFUL increase in cancer
occurrence, it is referred to the Environmental
Epidemiology Section for further investigation.
Subsequent investigations usually include detailed
case reviews and on-site interviews, and if necessary,
environmental sampling and monitoring of relevant
residential exposures. In some instances, a grant
proposal may be prepared to seek federal funds to
facilitate further study of the reported cluster.

RESULTS

Between September 1989 and December 1990,
fifty-two disease cluster reports have been received
from thirty-one counties throughout North Carolina.
Forty-nine of these reports related to cancer (Table
1 ) and three involved other health events (Table 2).
In addition, four re-evaluations of cluster studies
from prior years were conducted.

Of the forty-nine cancer cluster reports, the
Cancer Surveillance Section evaluated forty-three
and six reports were managed by others — one by the
Environmental Epidemiology Section; one by a local
physician from Washington County who reported a
potential cluster of colon cancer cases; two by the
epidemiologist in the Guilford County Health
Department; one by a University of North Carolina
researcher (Dr. Carl Shy); and the final one by a
graduate student from the Bowman-Gray School of
Medicine. Twenty-seven of these cluster reports
have been closed, eleven are still in active study and
another eleven are awaiting the start of active
evaluation ("Pending"). Of the closed reports, five
were determined to be REAL clusters of which three
were deemed to be MEANINGFUL. Each of these
five cluster evaluations are described below:

Brain Cancer in Rowan County — This repent
was suiTmitted fry a local ph-jsxcian who believed Ke had
seen an increase in the brain cancer cases in his medical
practice. Following this report, the Environmental
Epidemiology Section reviewed 260 medical histories (in
seven hospitab) over the next year which represented all
new brain caruxr cases for Rowan County arui its
neighboring counrie5 from 1980 to 1989. The case
histories include details of the tumors (dates of diagnosis,
cell types) arui any krwwn risk patterns (age, occupation,
srrwking history, etc.). Several cases initially identified by
the informant were excluded from the "cluster" because
the caru^er of the brain reported to the investigators was
spread to the brain from other body sites, or because the
person was diagnosed with brain career prior to moving to
Rowan County.

Statistical aiuilyses iiuiicated that an unusual pattern
of brain cancer incidence had occurred in Rowan County,
although not an increased rate of disease (Table 3). For
eight of the years of the eighties, very feu> brain cancer
cases occurred in the "cluster area." But in 1985 and
J 989, most of the Rowan County brain cancer cases
occurred in the ' 'cluster area. ' ' For the county as a whole,
however, no overall increase in brain cancer occurred.
The observed ' 'sudden ir^rease' ' actually followed periods
of low occurrence. This pattern of brain cancer occurrerux
did not suggest an environmental risk. The study results
were presented by written report and through a local press
conference.

Brain Tumors in 'Northampton County — This
report was submitted by the county health director who
was concerned about irurreased occurrerv:e of brain tumors
ar\d the possible association with a local industry. A
literature review iruiicated that the relationship to the
suspected industry was plausible aixd had been observed
by others (Rabotti, et al, 1966; Morantz, et al, 1985).
Community residents assisted health department personnel
in identifying living brain tumor cases; the Cancer
Surveiilance Section performed a parallel search for
deceased cases.

An inspection of the 1 984-89 observed-versus-expected
case pattern indicated eighteen brain tumors occurred
when twelve were expected. The race-sex pattern was not
coTuistent with the ruitional pattern run with the county's
population characteristics. A subset of four cases occurred
in proximity to the suspect iruiustry. The possibility that
these cases represented the increase could not be dismissed,
and as a result, this cluster was referred to the
Environmental Epidemiology Section for further study.

Pediatric Cancer in Gaston County — This
cluster report was submitted by the parent of one of three
cases of a rare pediatric tumor believed to have occurred
in a short period of time in a small municipality in Gaston
County. The search for additioruil cases from 1 970 to
1 989 iruiicated that there were two of these rare pediatric
caru:er cases in the mid-seventies followed by an eleven-
year hiatus and then six cases in the r\ext 30 months.
These latter cancer cases were the subject of the study. A
review of the scarce literature for this tumor revealed that
researchers in hlorth Carolirw. had previously identified
the iruyeased rates for this car\cer wthin a tu^elve-county
re^on {Crruffevman, et al. , 1 982).

The statistical aruilysis by the Cancer Surveillarux
Section iruiicated that the cases were not randomly
distributed. Sorrxe of the previously identified risk factors
for this carKer were also fourui with these cases. Since
these cases involved children, the distribution of birth
defects in Gaston County was also examined using data

from the N.C. Birth Defects Registry. No geographic
clustering was observed for birth defects. Despite this
firuiing, the presence of several large iruiustries and two
recognized hazardous waste sites in proximity to several of
the caivxr cases led to this cluster being referred to the
Environmental Epidemiology Section for further study.

Non-Hodgkin's Lym.phonia in Granville
County — A county commissioner referred this potential
caru:er cluster to the Cancer Surveillarxce Section. The
residents of this small community u«re quite cxmcemed
about an overall iru:rease in cancer, but especially Non-
Hodgkin's Lymphorrui (NHL). The aruilyses of the
Granville County mortality experience revealed that their
rates were rujt elevated over those of the state. Con-
sequently, further evaluation activities by the state would
ru}t be pursued. Upon hearing of these firuiings however,
local residents organized a citizen's task force to extend
the data aruxlyses further by collecting cancer irKideru:e
data for 1 975 to 1 989. This u^ell-organi^ed community
effort identified over 1 30 cancer cases, nine of which were
rum-Hodgkin's lymphomas. The citizen's task force also
collected emissions data for a dozen local iruiustries.

Using the citizen-identified cases, a REAL increase for
lymphoma occurrence was fourui, but there was ruj
particular spatial or temporal pattern for the cases. There
was abo a visually identifiable aggregate of colon caruxr
cases, but overall there was no iru:rease in colon caruxr
occurrence. Norxe of the cancer patterns could be linked to
any of the paths of iruiustrial emissUms; arui only one
company was determined to represent a potential hazard.
This cluster report was closed with the assurarxce that
surveillaru:e would be maintair\ed for this community.
While the work of the citizen's task force did not change
the firuiings regarding cancer clustering, it did provide
better data for the aruilysis arui a further opportunity for
education to the community about cancer.

Colon Cancer in New Hanover County — The

county health director reported this cluster of four
intestinal caruxrs arrumg neighbors, living in extremely
close f>roximity. Three of the four colon car\cer cases were
urxder 65 years of age, an unusual distribution for caruxr
at this aruitomic site. This region of the state has been the
focus of much study because of drinking water con-
tamination (e.g., trihalometharie formation). Eachofthe
residents' drinking water was taken from a private well
which was suspected to be contamiruited with hydrogen
sulfide. This cluster report was referred to the
Environmental Epidemiology Section for testing of the
drinking water supplies. In light of the proximity of
another "closed" cluster report to this one, a
recommeruiation was also made that surveillarux be
nmintained for cancer occurrerux in both of these
communities.

In addirion to the study of cancer clusters, the
Cancer Surveillance Section and the Environmental
Epidemiology Section perform special studies and
apply for federal grants to fund large research efforts
in North Carolina. A special study of brain cancer in
North Carolina is in progress. The information from
both the Rowan and the Northampton clusters
figures significantly in the design of that research
effort. A federal grant proposal is being prepared to
conduct a more thorough study of the several
counties potentially involved in the rare pediatric
cancer cluster in Gaston County and of the
environmental hazards identified through that
investigation. Finally, a large study of leukemia
patterns in the state is also under way; this is
associated with findings from several cluster report
evaluations.

FOLLOW-UP OF CLOSED CANCER
CLUSTER REPORTS

With the additional capabilities that are now
available for the statistical analysis of cancer cluster
reports, and because of the continuing interest
expressed by three of the communities that had
reported earlier cancer cluster reports, three evalua-
tions for cancer clustering were re-analyzed for this
report. Also, a search for evidence of leukemia
clustering was repeated due to continuing interest by
the Environmental Epidemiology Section.

Leukemia in Cherokee County — Local residents
suspected increased cancer occurrence due to the aerial
spraying of herbicides in the forests of western North
Carolina. Researchers evaluated the pattern of mortality
for Cherokee County and seven surrounding counties
using statistical tests. No pattern of increased cancer
occurrence u>as found to be associated udth any potential
environmental hazard. TKe finding of a generalized
increase in all cancers for two counties (Cherokee and
}sAacon) led to a recommendation of further study, with
an exparxded time period.

The Cancer Surveillance Section re-analyzed the
leukemia data for these eight counties arui extended the
aruilyses up to 1988 (Table 4). Macon Counfy had two
time periods that showed evidence of temporal clustering,
1979-81 and 1987-88. For Cherokee and Haywood
counties as well as the region as a whole, the years
1980-81 also showed increased leukemia rrwrtality. It is
intriguing that the cancer occunerice for the years 1 980-
81 of the previous study was the basis upon which a
follow-up aruilysis was recommeruied; yet these very years
(1980-81) may rww be seen to represent the peak of
leukemia occurrerice for the region. It will be interesting to
monitor future cancer patterns for these counties.

Leukemia ai\d Lymphoma in Burke County —

Researchers from the Vnwersity of North Carolina
corxducteda "rxearest-rKighbor" arwdysis of several caruxr
cases in Burke County for the period 1970-79. Their
results did not indicate clustering for leukemia or
lymphomas as reported by citizens. Clustering of lung and
prostate carxcers was found, however. The CDC also
evaluated this leukemia /lymphorrux pattern and/ound no
evidence of increased occurreru:e.

The Cancer Surveillance Section analysed the
lymphorrwL iruidence data for Burke County for the 1 980-
89 period. There is still no eviderxce of iruyeased
occurrerxce, hut the ten-year pattern is interesting. Of the
fourteen cases in these ten years, eight occurred in 1 980-
82 (three, three and two cases respectively). Two of the
1 980 cases lived close to each other, but none of the others
lived particularly close, nor was there any unusual age,
race or sex pattern evident. The fact that two cases lived
near the location of the earlier suspected cluster is also
notable. Further surveillaru:e of the leukemia and
lymphoma patterrxs in Burke County is in order. An
evaluation has been completed for lung carvxr in Burke
County and no irxcreased occurrence was found.

Cancer in Davie County — A request was swfcmitted
to the Environmental Epidemiology Section in J 986 for a
study of suspected higher-than-expected cancer rates in
Davie County. Residents u^ere particularly concerned
about hazardous emissions from irxdustries near the town
of Cooleemee. The community organized a survey to
identify additiorml cases; organizatUmal efforts were so
successful that a unit of the American Cancer Society was
formed from that beginning. Hou^ever, the conclusion
from this study was that there was no evidence of
increased carxcer rates.

Using the CLUSTER software, an arxalysis of the
cancer experience in Davie Courxiy was re-evaluated arxd
extended through 1 988. Over the J 984-88 period, Davie
County's liver and bladder cancer death rates were
elevated above the state rates; however these excesses were
based on very srruxll nurrxbers (i.e., averages oforxe death
or less per year) and therefore are not considered reliable
estimates. The Davie County lung, leukemia, kidrxey and
brain cancer death rates were lower than the state's.

Myelogenous Leukemia in "North Carolirui —

The Environmental Epidemiology Section requested an
arwdysis of 1984-88 age-adjxisted mortality rates of
myelogerxous leukemia. Their objective was a search for
potential clusters associated with point sources ofairborrxe
emissions ofbenzerxe. A visiml irxspection of a map with
the myelogerxous leukemia rates wets performed at the tinxe
of the initial request. Now a statistical arxalysis for
evidence of clustering is available using the CLUSTER
software (see Figure 3).

Twenty-six counties had very high myelogenous leukemia
mortality rates (> 3.4/100,000); some of these rates in
extreme eastern and western counties may he urxstable due
to the very small numbers oj cases involved. The state ar\d
ruitioruxl rates of myelogenous leukemia mortality were
2.5/100,000. Rates for forty-four counties were
significantlyhigher (i.e., ^ 2.7/ 100,000). The distribution
of these counties with significantly elevated rates provides
evidence of several clusters that may offer leads for further
investigation (probability of observed clusters p < 0.001;
p < 0.05 for counties with very high rates). A special
study of leukemia occurrence in the central Piedmont
counties is already under way.

DISCUSSION

Many public health experts believe that evaluating
cancer cluster reports is a valuable service. The new
protocol developed for evaluating cancer cluster
reports has greatly increased the responsiveness to
citizen concerns, e.g., twenty-seven cluster reports
were "closed" using the new protocol in the
comparable time that one report (Rowan County)
was evaluated using former procedures. The new
protocol already provides greater cost-effectiveness
and timeliness in handling cancer cluster reports
through the use of statistical analyses of the available
data in place of the costly process of collecting new
data for each cluster. As the new protocol is refined,
it will provide a systematic approach to cancer
cluster investigation and analysis; this will mean a
straightforward, step-by-step process that can be
explained to persons reporting clusters.

Also, much greater contact is being made with
local residents to provide educational programs
(nine on-site lectures were provided as a part of the
twenty-seven "closed" cancer cluster evaluations).
These community presentations help to assuage
community concerns and extend a positive response
to public inquiries. These educational programs
provide answers to legitimate questions about both
cancer prevalence and possible environmental
associations. To facilitate these educational goals, an
informational brochure was developed in collabora-
tion with the American Cancer Society; it is available
by request. In addition to the evaluations described
in this report, the staff of the Cancer Surveillance

Section have been involved in four consultations
related to potential cancer clusters in North Carolina,
one in each of Harnett, Robeson, Burke and Catawba
counties.

This greater effectiveness and visibility have led to
reports of disease clusters other than cancer being
referred to the Cancer Surveillance Section. Three
such cluster reports are:

A potential cluster of Multiple Sclerosis in
Alleghany County.

Two cases of a rare dermatologic condition
diagnosed in neighbors in Cabarrus County.

A series of spontaneous abortions, in a two-
month period, among neighbors in Wake
County.

However, for so many cluster reports, even the
time required for site visits and data analyses compels
the Cancer Surveillance Section to do cluster
evaluations as time permits and where travel makes
site visits convenient. The work with cancer clusters
will continue, although recent budget constraints
will lengthen the time to complete evaluations. The
twenty-seven reports described here as "closed"
averaged three months for evaluation; this does not
include the extensive Rowan study that required 14
months for completion. Such rigorous and detailed
efforts must be reserved for those clusters that truly
warrant such intensity, including those identified
through proactive cancer surveillance.

CONCLUSION

Cancer cluster investigations may represent a
social service more than scientific research. Cluster
investigations should not be viewed strictly for their
scientific research merit, but also for their role as an
important interactive public health activity. "The
public is increasingly demanding answers regarding
possible associations between [cancer] and the
environment." (Fiore, et al., 1990). The evolving
procedures described in this report address these
concerns with a sound epidemiologic approach to
responding to cancer cluster reports.

REFERENCES

Aldrich, T.E., Atkinson, D.A., Hines, A. and Smith, C.G. ( 1990), The Establishment of a Population-Based
Cancer Registry for North Carolina, North Carolina Medical Journal, 5 1(2): 107-12.

Caldwell, G.G. (1989) Time-Space Cancer Clusters, HealtK and Environment Digest, 3(5): 1-3.

Fiore, B.J., Haranhan, L.P. and Anderson, H. A. (1990), State Health Department Response to Disease Cluster
Reports: A Protocol for Investigation, American Journal of Epidemiology, 132:S14-S22.

Grufferman, S., Wang, H.H., Delong, E.R., et al. (1982), Environmental Factors in the Etiology of
Rhabdomyosarcoma in Childhood, J. Nat. Cancer Inst., 68 (1): 107-13.

Morant2, R.A., Neuberger, J.S., Baker, L.H., et al. ( 1985), Epidemiologic Findings in a Brain-Tumor Cluster in
Western Missouri, J. Neurosurgery, 62: 856-60.

Rabotti, G.F., Grove, A.S., Sellars, R.L., and Anderson, W.R. (1966), Induction of Multiple Brain Tumors in
Dogs by Rous Sarcoma Virus, Nature, 20:884-6.

SEER (1990), Cancer Statistics Review 1973-87, Pub. Surveillance Program, National Cancer Institute, NIH
Pub. No. 90-2789.

Thacker, S. (1989), Time-Space Clusters: The Public Health Dilemma, HealtK and Environvxent Digest,
3(5):4-5.

Warner, S.S. and Aldrich, T.E. (1988), The Status of Cancer Cluster Investigations Undertaken by State
Health Departments and the Development of A Standard Approach, J. American Public Health Assoc.,
78(3):306-7.

Selected Bibliography for Additional Reading

National Conference On Clustering of Health Events, American Journal of Epidemiology (1990), 132 (1)
Supplement, Twenty-six papers. Available from the Journal office, 2007 E. Monument Street, Baltimore, MD
21205 for $11.00.

Guidelines for Investigating Clusters of Health Events, Morbidit^i and Mortality Weekl:y Report ( 1990), Vol. 39
(RR-11), 23 pages, 35 references. Available from the Massachusetts Medical Society, C.S.P.O. Box 9120,
Waltham, MA 02254-9120 for $3.00.

Aldrich, T.E. (1990), CLUSTER: Software to Assist Investigations of Rare Health Events, Diskettes and
User's Manual — Available from the Agency for Toxic Substances, Health Studies Division, MS-E31, 1600
Clifton Road NE, Atlanta, GA 30333 (no price identified).

RIPOAT OF
POSSIBLE
DISEASE
CLUSTER

PROXIPML DATA
RCVIEV

CONTACT WITH
\ LOCAL ,.
^ACEUCY NTT-
f.

REPORTING

FILE ^
KAIHTAIHEMCE^

CONTINUED
SURVEILLANCE

PREVENTION

DISCONTINUE
INVESTIGATION
/

r^.

NO EVIDENCE OF
auSTERING

ON-SITE J r

INTERVI ElC^y^NON-CASES

CON FOUNDERS

I I

I
QRS^

STATISTICAL
ANALYSIS

RISK
FACTi

PROBABLE
EXPOSURES

. PRimRY
-^ EVALUATION -
OF PROPOSED
CLUSTER

.DECISION POINT

INCONCLUSIVE
EVIDENCE OF
CLUSTERING

EHCNDCO
EPIDEM10L06K
STUDY

EXTENSIVE
EXPOSURE
ASSESSMENT

POSSIBLE
CLUSTERING

Figure 1 — North Carolina Protocol for Evaluating Cancer Cluster Reports

Cancer Cluster Reports

Received By County

September 1989 Through December 1990

N.C. Cluster Reports

no cluster reports received [H

1 cluster report received Fl

2-3 cluster reports received H

4 or more reports ■

Figure 2 — Cancer Cluster Reports by County 9/1/89 to 12/31/90

Age- Race- Sex Adjusted Death Rates
for Myelogenous Leukemia

North Carolina, 1984-88

Deaths per 100,000
Population

D 0.00 to 2.49

D 2.50 to 2.69

M 2.70 to 3.39

■ 3.40 to 8.00

(1970 U.S. standard)

* Rates of 2.7 or greater were significant at .05 level

Figure 3 — Map of Myelogenous Leukemia Mortality 1984-1988
Age- Race- Sex-adjusted Rates (1970 U.S. standard)

Table 1 — North Carolina Cancer Cluster Reports Received 9/1/89 to 12/31/90

County

Alleghany

Ashe

Beaufort

rWaufort

Bertie

Brunswick

Burke

Caldwell

Catawba

Catawba
Chowan
Cleveland

Cumberland

Davidson

Forsyth

Gaston
Gaston
Gaston

Granville
Guilford
Guilford

Johnston

Mecklenburg

Mecklenburg
Mecklenburg
Mecklenburg

New Hanover
New Hanover
Northampton

Rockingham
Rockingham
Rowan

Rutherford

Stanly

Surry
Surry
Surry

Union
Wake
Wake

Wake
Wake
Warren

Washington
Washington
Wayne

Wilson

Description

Possible excessive rates in small town
Concern for excess cancer in the county*
Suspected drinking water contamination

Concern for excess cancers in local area
Possible increased cancer among neighbors
Brain cancer around nuclear power plant

Fear of lung cancer due to plant emissions*
Kidney cancer cases in small area
Concern for cancers near industrial park

Six neighborhood cancers in two years
Concern for local paper plant emissiotis
Multiple cases in one neighborhood*

Eight cancer cases among 30 homes
Concern for cancers near industrial park
Four gastrointestinal cancers in neighborhood

Three rare pediatric tumors in small town*
Concern for cancers near industrial park
Hodgkin's Disease in rural area

Concern for increased lymphomas*

Possible excess rates in Summerfield community

Possible excess rates in Jamestown neighborhood

Concern for increased cancer in neighborhood
Concern for increased cancer in neighborhood
Three leukemia cases in small area

Concern for increased cancer in neighborhood*

Concern for industrial emissions

Six breast cancer cases in neighborhood

Four colon cancer cases in neighborhood