EPA 560/11-80-013
March 1980
LUNG CANCER/MORTALITY IN PROXIMITY
TO A
PESTICIDE PLANT
Genevieve M, Matanoski*
Emanuel Landau0
James Tonascia*
Christiana Lazar*
Elizabeth A. Elliott*
William McEnore*
Richard Keating*
Joseph Seifter+
.Johns Hopkins University*
School of Hygiene and Public Health
Baltimore, Maryland
American Public Health Association0
Washington, D.C.
U.S. Environmental Protection Agency+
Office of Toxic Substances
Washington, D.C.
Project Officer
Charles Poole
Office of Toxic Substances
Washington, D.C.
-------
DISCLAIMER
This project has been funded with Federal funds from the
Environmental Protection Agency under contract number
68-01-3859. The content of this publication does not
necessarily reflect the views or policies of the U.S.
Environmental Protection Agency, nor does mention of trade
names, commercial products, or organizations imply
endorsement by the U.S. Government.
-------
ABSTRACT
This interim report covers a study of excess mortality from
cancer in the population residing near a chemical plant
which had produced insecticides, herbicides and other
arsenic products for three quarters of a century in the
inner city of Baltimore.
A pilot study of three years mortality in the area compared
to the remainder of the city indicated a three fold excess
of lung cancer mortality for white males. As a result,
cancer deaths were examined for a five and nine year period
surrounding census years to determine not only the absolute
mortality but the changing trends in approximately a twenty-
year interval as part of the larger study.
Four index census tracts including the one containing the
plant were selected based on the requirement that at least
50 percent of the area lay within a three-quarter mile
radius of the plant. Comparison tracts were selected based
on matching the index tracts by age, race, sex, median
income, and for 1970 only, percent poverty level and percent
head of household over age 65. Deaths among employees of
the plant were obtained and subtracted to diminish the
effect of possible direct occupational exposure to arsenic.
The lung cancer death rate on an age adjusted basis has been
found to be significantly higher in the census tract containing
the plant compared with other index tracts and controls. It
has been rising rapidly since the mid-sixties. A study of
the validation of hospital records and the pathology of
cancer cases indicates that the excess risk of lung cancer
represents a real risk in mortality and is not the result of
local diagnostic and certification practices.
A corollary study of soil arsenic levels indicated that the
recent contamination was less than previously experienced
since the values at two inches below the surface were higher
than at the surface in most areas. The pattern of high soil
levels appeared to be related to rail transport of the
arsenical material. High levels of arsenic in a generally
northerly pattern were correlated with the location of the
highest frequency of lung cancer cases as seen on the spot
map.
Further research is required to determine whether smoking
patterns, non-pesticide associated occupational exposure
or residential history may play a role in the excess of lung
cancer.
-------
LUNG CANCER MORTALITY
IN
PROXIMITY TO A PESTICIDE PLANT
GENEVIEVE M. MATANOSKI, EMANUEL LANDAU,
JAMES -A.. TONASCIA, CHRISTIANA LAZAR,
ELIZABETH A. ELLIOTT, WILLIAM McENROE,
RICHARD KEATING, AND JOSEPH SEIFTER
Introduction
Arsenic has long been recognized as a poison when ingested
in large quantities by man, animals or plants. It is known
that continued ingestion of high natural levels of arsenic
in water or food .will produce skin lesions including cancer
(1,2). Consumption of arsenic as a therapeutic agent is
also known to cause skin lesions (3).
The risk of inhalation of arsenic has not been as extensively
investigated. Workers exposed to arsenic in the manufacture
of pesticides have an increased risk.of lung cancer and
possibly lymphomas (4,5,6). Less is known about the chronic
health effects in the general population exposed to arsenic
in the air. Children around smelters may have high arsenic
levels in nails and hair but it is not clear whether these
observed indications of absorption of the agent can also
produce long-term toxicity. Blot and Fraumeni (7) have
-------
suggested that there may be an association between excessive
lung cancer mortality and the existence of non-ferrous
smelting industries in several counties in the U.S. It is
possible that some by-product of this industry such as
arsenic is associated with these carcinogenic effects.
The purpose of the current study was to determine whether
there is an excess mortality from cancer in the population
which resides near a chemical plant in the inner city of
Baltimore and whether any observed excess can be associated
with previous exposure to arsenic. The plant has produced
insecticides, herbicides, and other arsenic products from
1897 until early 1976. In 1952, the original plant was torn
down and a new one erected with better hygienic conditions
for the workers. The plant produced arsenic acid, calcium
and lead arsenate, Paris green (a cupric acetoarsenite), and
sodium arsenite. All products were dried and packaged
except sodium arsenite which was shipped as a liquid. Paris
green was not produced after the early 1950's and no dry
arsenicals were produced after 1973. Other pesticides such
as chlorinated hydrocarbons and organophosphates were not
produced at this facility but were made into formulations
on-site since 1947. Further information about plant operations
has not been sought.
There are several other industries which are currently
-------
located in the area or have been manufacturing in that
vicinity in the past. Adjacent to thechemical plant is a
utilities company where natural gas was produced in the past
through either a carburetted water gas or later an oil gas
process. Currently, the company only stores natural gas in
that area. Railroad yards are located to the northwest and
southeast of the plant. There are other manufacturing
industries in the area at this time and several have existed
in the past within a one mile radius of the plant. These
included a company which made rubber products, a small plant
which produced fishing nets through a process which required
dipping the product into organic materials and subsequent
drying, and an abbatoir. Most of these operations have been
defunct for the past ten years and little is known about
their potential environmental hazards during the time they
existed.
The pilot study, completed in April 1976, compared the
three-year cancer mortality in this area to the rest of
Baltimore. The results indicated that the census tract in
which the plant was located had a higher all-cancer death
rate than the rest of the city. 8/ The mortality for lung
cancer specifically was three times higher in white males
from that area than in men residing in similar areas throughout
the city. The proposed expanded study of mortality was to
include cancer deaths for each 5 year period surrounding a
census year in order to determine not only the absolute
-------
mortality but the changing trends over three census periods
and to discover the possible errors in rates which resulted
from deaths outside the city.
Method
The census tracts which were selected as having had possible
environmental exposure to arsenic from the point source of
the pesticide plant were defined empirically as those for
which at least 50 percent of their area lay within a 3/4
mile radius of the plant. This distance was chosen so that
large tracts which lay across the river and in which the
majority of the population did not reside within a one mile
radius of the plant would not be included. The four index
census tracts which fit these criteria were 2303, 2302, 2404
and 2301. The tract in which the plant was located was 2303.
The comparison group of census tracts consisted of all
tracts which matched the index ones on age distribution,
race, sex, median income, and, for 1970 only, percent
poverty level and percent head of household over age 65
years. Two separate matching groups were selected. Those
designated as Match I were chosen for their similarity to
the three tracts 2303, 2302, and 2404 which were alike
especially in age and race distribution. The index tract
2301 had to be matched separately because its characteristics
particularly in racial distribution, varied from the other
-------
index tracts. By using this procedure for each census
period, we found that different tracts matched in different
census years. We also discovered that the boundaries of
census tracts differed from one time to another. The most
marked changes were between the characteristics of census
tracts in 1950 and 1970, the extremes of time. However, we
have utilized the same criteria for matching and have
selected tracts in all time periods in the same way. Data
from 1950-51 have not been analyzed for presentation at this
time. For deaths in 1958 through 1962, the 1960 census
tract data were used both for calculation of rates and
establishment of matching tracts. For all other years, the
1970 census data were used for these purposes. Using
separate matching procedures for each time period assured
that for each period, the index and control tracts would be
comparable for those variables on which we matched and
mortality could be directly compared. However, they were
not necessarily similar in matching' variables over time.
This did not concern us for race and sex which were analyzed
separately but the age distribution of the populations in
the index and control tracts had changed slightly. To
account for this, some tables include an age-adjustment
especially when rates are compared over time.
The initial matching criteria for the 1970-72 period were:
-------
Age distribution -f 10% for each age
Race +15% expressed as % black
Sex +_ 5% expressed as % male
Median Income +_ $1,000
% below poverty level +10% in 1970 only
% head of household .over 65 years + 20% in 1970 only.
These figures for ranges in matching were arrived at empirically
although a preliminary review of nearby similar tracts had
suggested that these values would include the adjacent areas
as controls. Thereafter, all tracts in the city which fell
within these ranges were included as controls. We realized
that using absolute income differences would represent a
different extent of variation depending on the relative
worth of the dollar in each decade. Therefore, we have
expressed the original dollar difference of $1,000 as a
percent of the median income in index tracts and this value
of 12% was used to determine the differences in median
income in index and control tracts in each census period.
In 1960, this represented $700 and in 1950, $400 as maximum
allowable differences in median income for matching tracts.
This method of calculating relative income was based on the
assumption of economic stability in the index area. Similar
values would have been attained, however, if we had assumed
a five percent annual income increase over the 10-year
period. The matching tracts in 1970 and 1960 are shown in
Tables 1 and 2. A total of 18 tracts matched the three
-------
similar index tracts in the 1970 census and a total of 5
tracts match the 50-50 racial distribution tract in both
the 1970 and 1960 censuses and these are designated as Match
II. The variation in the number of matching tracts in the
two calendar time periods is a reflection of the level of
racial integration which has taken place in some of the
middle income tracts over the past ten years. Those integrated
tracts are no longer comparable in racial mix to the index
tracts and so have dropped out in the 1970 matches. Also,
housing in some central tracts was demolished during inner
city reconstruction.
The comparabilities of index and control tracts are indicated
in Tables 3 and 4 which display the characteristics of the
tracts for four of the matching variables. As can be seen
from these tables, the proportion of individuals aged 45
years and over residing in the index tracts has increased in
the last decade. This is an indication of the stability of
the population in the area near the plant. As a further
marker of the permanence of the population, Table 5 indicates
the mean percentage of the population occupying the household
for seven or more years in 1960 and five plus years in 1970.
These indicators would suggest that the Match I tracts and
controls are very similar in population stability. Match II
controls and index tract 2301 have more raobile populations.
The tract 2303 which contains the plant has increased the
-------
percentage of long-term residents from 55 in 1960 to 67 in
1970. It seems reasonable to assume that most of the 55
percent population which had been in tract 2303 since 1953
or earlier (residents seven or more years) have not all left
the area in the next 5 years but constitute a high proportion
of the 67 percent which were known to have occupied their
homes from 1965 or earlier during the 1970 census (residents
five or more years). It is highly probable, therefore, that
over half of the population in that tract had lived in the
same house for 17 or more years by the time of the 1970
census. For this reason, deaths occurring in residents
during the period from 1968 on would probably represent the
experience of long-term residents of the area. This explanation
is only a presumption and could not be proven without a
survey of the population to determine residence histories of
households. We are hoping to be able to make this necessary
survey.
All the previous tables have subdivided the control into
four groups, the adjacent, the southern, the northern and
the central controls. Figures 1 and 2 indicate the geographic
distribution of all census tracts for both matching groups.
The scattered distribution of the control tracts has placed
them in areas which may also have had different risks. The
adjacent tracts contiguous to the index ones may have had
minimal exposure to the same agents in the major area.
-------
Southern tracts are in heavy industrial areas as are the
central tracts but the characteristics of the populations
and their stability are different. The northern area
consists of mainly residential dwellings with little industrial
exposure.
Cancers were identified by examining all certificates of
deaths which occurred within the city. The death was
selected for study if cancer appeared as a cause listed
anywhere on the death certificate. Deaths of city residents
were selected from the total cancer list. This procedure
would not include the deaths of city residents which occurred
outside the city. These deaths had to be identified from
another source which would have required more funding and
time than originally proposed. If residents of certain
areas of the city were more likely to have died at county
hospitals than residents of other areas, we might have
under-estimated the deaths to a greater extent in one area
than the other and thus created a difference in mortality.
In order to determine the extent of these differences we
abstracted information on out-of-city deaths of city residents
for the three years 1970-72. The proportional increase in
deaths for the index tracts was 5 percent and for the
control tracts 13 to 15 percent. This difference is not
large enough to account for the variation in cancer rates
observed. .
-------
Deaths were included only once regardless of number of
cancer causes listed on a death certificate. The death was
attributed to the underlying cancer or to the first cancer
listed if there was an underlying cause of death other than
cancer. Data from all census tracts were collected and
analyzed in the same way so that comparisons between the
matched tracts as well as with the total city are valid.
The original design of the study was altered to include more
years around the 1970 census when the most rapid changes in
cancer rates were observed. For the years 1966-67, the
procedure described above was used since other data available
for 1967 had missing deaths. However, for 1973-74 only the
underlying cause of death as listed on a print-out has been
used to identify cancers since death certificates were not
readily available. Very few cancers, especially those in
the lung, are missed by using this procedure.
The completeness of ascertainment of cancer deaths was
determined by checking all-records with a second abstractor.
The omissions were 1.1 percent for the first abstraction.
About half of the omissions were in cancers listed as
underlying. The errors were corrected so that the overall
missed cases after two reviews should be very small. The
completeness of identification was also validated subsequently
by matching our deaths against a tape of all deaths listed
by underlying cause for all residents between 1968-72.
Essentially no additional deaths were found by this method.
-------
The coding of causes of. death used two different classification
schemes in the two periods with the period around 1960 using
the 7th revision and the period of 1970, the 8th version.
We examined the differences that these coding revisions had
made in the classification of deaths and found that for some
cancers the ICDA codes were essentially unchanged. We have
included lung cancer not specified as primary within the
total group of lung cancers. This procedure was indicated
on the basis of coding practices and differences in the
methods of recording causes of deaths in different census
tracts. If "cancer of the lung" was recorded on a death
certificate even though that was the only cause listed, it
was usually coded as if it was not a primary lesion. The
condition had to be specified as primary to be classified in
that category. We found that the proportion of lung cancers,
especially in the 7th revision, designated as primary
differed by census tract and by time period even though the
certificates appeared .to be the same. We felt that the
proportion of lung cancer deaths which were attributed to a
primary lesion might depend on the precision with which
local doctors filled out death certificates as well as
coding practices, and thus, in turn, might vary by census
tract and time. To combine the two classifications should
not include the metastatic lesions but would include all
types of lung cancers. For cancers such as cervix, breast,
bladder, and prostate in which the code is entirely different
-------
in the two coding revisions, the appropriate code for each
time period was included separately.
Results
Mortality by Tracts
The crude rates for cancers at four specific sites-oral,
pancreas, lung and prostate as well as for all cancers are
presented for males in Tables 6 and 7. The former table
includes data for the five-year period around the 1960
census and the latter table for a seven-year period around
the 1970 census. The rates for individual tracts are examined
separately since the tracts are located at different distances
from the plant and therefore could have had a gradient in
the populations exposure to pesticides. As can be seen, the
risk for lung cancer and for all cancers is excessive in the
period around the 1970 census for tract 2303 compared to any
of the control groups. This is not consistently true in the
earlier period. The weighted relative risk of lung cancer
in white males from index tract 2303 as compared to north
controls which had the highest control rate is 2.5 shown in
Table 7 with a probability of .0005 as determined by the chi
square calculated by the Woolf-Haldane method (9,10). The
black males in tract 2301 also have a higher rate of lung
cancer but this was not true for white males in the same
tract. In the 1960 census period, although the lung cancer
-------
rate is higher in white males in tract 2303 than the north
and south controls, there is very little difference between
the rates for all index tracts and for the adjacent and
central controls. There are no differences in rates for
males in tract 2301 and their control groups. Since the
index and control tracts were matched for several variables
no attempt was made to adjust for age in the initial calculations
of mortality rates. Where differences have been noted in
the first comparisons, adjusted rates will be presented
subsequently.
For most control groups, the census tracts included in two
census periods were not the same. However, some of the
central area tracts from the 1960 and 1970 censuses were the
same and rates are presented in Table 7A for those control
and index tracts which are the same in the two periods. The
rates for the central tracts are approximately the same for
1960 whether or not all tracts or the identical tracts from
both censuses are included. These data would suggest that
the rates for all of the areas were relatively similar in
1958-62 and that only 2303 had experienced a four-fold
increase in lung cancer mortality and a two-fold increase in
all'cancers-over the following 10-year period.
If we examine the comparable crude mortality rates for
females in Tables 8 and 9, we can find no significant excess
-------
risk of cancer at any site for census area 2303, In fact,
the overall cancer rate appears somewhat low especially in
the 1958-62 period. The mortality from breast cancer is
slightly high in the early period and there are no deaths
from cervical cancers. The lack of an observed increase in
lung cancer in women in 2303 might be the result of a small
population size. This will be discussed later in the
report.
Mortalityby Time
It was necessary to determine more precisely the time at
which the increased mortality from lung cancer began in
census tract 2303. Table 10 lists the crude average annual
mortality rates for lung cancer in males for the two
five-year periods around the census years and for two
two-year periods before and after the 1968-72 rate. After
the initial five-year period, 1958-62, when the mortality
from lung cancer is similar for all index tracts and for the
adjacent and central control tracts, the rate in tract 2303
shows only a slight rise in 1966-67. Throughout all of this
early period, however, the lung cancer mortality is significantly
higher in 2303 than in the northern controls. By 1968-72
the lung cancer mortality in males in 2303 has risen to
about 4 times its original rate whereas the other areas
shown have either a much less dramatic increase or no
increase at all. The excess risk has persisted to 1973-74.
-------
Both periods have significantly higher rates for 2303 as
compared to northern controls. The rates for tract 2301
which lies north of the plant have also increased in this
time period but the change is less impressive. All other
index tracts have had a constant or decreasing rate.
A d j u s t e d - r a t e s
The age distribution of the index tract changed with time
and these differences were reflected in similar changes in
the control group. In order to have appropriate comparisons
over time, the mortality rates for each cancer site and for
all cancers have been adjusted using the method of standardized
mortality ratios. The average annual Baltimore City mortality
rates were calculated from all deaths in the 1968-74 period
and these values were used as standards to adjust the
mortality in each time period. As seen in Table 11, the
mortality ratios for white males in tract 2303 were high for
cancers of the lung, pancreas, stomach, prostate, oral
cavity and all sites. The numbers of deaths except for lung
and all sites combined were small but the pancreas cancer
rate was still significantly higher than that for the city.
White females in 2303 had an unremarkable overall cancer
rate with excesses noted only for oral and rectal cancers of
which only the latter ratio is significantly greater than
unity. Since the colon cancer ratio is so low, it may be
that an artifact of diagnosis locally may be classifying
-------
colon cancers as rectal cancers erroneously. It seems
unusual, however, that this should be happening only in
females.
In Tables 12 and 13 and Figures 3 and 4 we examine the lung
cancer mortality in two or three year time intervals. Using
rates adjusted by the direct method to the 1970 Baltimore
City population as a standard, we find that the death rate
for this cancer has always been higher in males from tract
2303 than from most controls but that it has been rising
rapidly. The total cancer rates have also shown higher
values than among controls. A preliminary look at the lung
cancer rates for 1950-51 indicated that the adjusted rates
for that period were high for tract 2303 with a rate of 253
per 100,000 population as compared to rates ranging from
35.8 to 87.4 in other index tracts and controls.
Employees of an industry may live in close proximity to
their place of work and could have accounted for an increased
mortality in the census tract due to occupational exposure
in the pesticide plant. With the cooperation of Dr. Abraham
Lilienfeld of Johns Hopkins University and the company we
reviewed lists of all employees to match with known deaths.
Four employees were found among the cancer deaths in tract
2303 and two of these occurred among the lung cancers
reported in 1969 and 1970. We felt it was important to
determine how the index tract 2303 compared to all others in
-------
the city and not just those tracts selected for study. The
standardized mortality rations for lung cancer deaths in
white males in every census tract in the city have been
compared to the total white male lung cancer rate for
Baltimore for the period 1970-72.
In table 12A deaths among employees for the index and
adjacent tracts have been removed from the numerator and
denominator of the rates. The resulting value is a minimum
rate since we did not include any record of which employees
were alive and residing in these same areas during the study
period and thus, could not withdraw the living individuals
from the denominator. The resulting rates still suggest
that 2303 has a high mortality from lung cancer although
1968-69 rates now are lower than the northern controls.
Only tracts with SMR's of two or more have been included in
Table 14 and Figure 5. These represent 14 tracts out of the
total of 201 populated census tracts in the city. Tract
2303 has the highest ratio, the value being 4.21 times
higher than the overall city rate. The nearest ratio of any
other tract is 2.78.
The geographic distribution of cases was plotted on spot
maps as shown in Figures 6 and 7 for two time periods. For
both periods, lung cancer appears to be concentrated in an
area about eight blocks wide and nine to twelve blocks long
-------
lying to the north and east of the plant. The area encompasses
all of tract 2303 and parts of 2302 and 2301. If one takes
all of tracts 2303, 2301 and 2302 which lie within a 3/4
mile radius of the plant and calculates the proportion of
lung cancers to all cancers in this area compared to the
remaining census areas shown in Figure 6, the proportion is
44.4 percent near the plant and 16.8 percent in the outer
areas. For Figure 7, the proportions show similar differences
for 1973-74 as in the previous three years. For the area
within the defined census tract and 3/4 miles from the
plant, lung cancer represents 47.8 percent of the total
cancers whereas in other areas it is only 33.3 percent. The
northerly direction of this lung cancer excess is not
compatible with the strong wind directions in that area.
These winds arise in the northwest and west and should have
carried contaminants to the east and southeast of the plant.
The particles may have been moved by gentler winds and
deposited nearby. (See Figure 8)
Piscuss ion
An excess mortality from lung cancer has been demonstrated
among men in a highly industrialized area of South Baltimore
over the period from 1966 through 1974. Although the study
was initiated to determine whether there was an environmental
impact from exposure to arsenic, the current information
does not provide evidence of a direct association.
-------
There are some questions which arise with regard to the
data. Why did the excess risk appear primarily in the late
60's and early 70's when the plant had existed and produced
arsenical products since the early 1900's? The discrepancy
could indicate that the plant did not account for the excess
but some other local industry or occupational group accounts
for the excess. It is also possible that the men in the
area had a higher frequency of smoking and smoked more
cigarettes than did the populations in the rest of the city.
Moreover, it is possible that selective out-migration of
younger, healthier males created a higher risk among the ;
remaining population. Both of the latter explanations could
be determined by a survey.
The sudden rise in lung cancer might be related to the
destruction of the old plant in 1952. Such an undertaking
could have spread dust diffusely throughout the community.
Under these circumstances, we must ask why the concentration
of lung cancer in the area does not coincide with the
assumed wind spread of particles. It is necessary to
examine further the mortality in the 1950 period to determine
whether an excess existed at that time before the destruction
of the old plant. It would be interesting to see if the
appearance of the excess risk .of lung cancer in the community
coincided with that found in the workers within the pesticide
plant. If we presume that arsenic may not be causing the
excess, then it would be necessary to examine the mortality
experience of workers in other industries in the area,
-------
especially the natural gas plant, to see if they have an
excess lung cancer mortality. In almost all cases, workers
within an industry should have higher exposure and a greater
risk of disease than the general public. It is necessary to
investigate whether the increase in lung cancer can be
related to a change in production or methods of operation of
any of the businesses. For example, differences in handling
arsenic, changes in formulation of pesticides or the conversion
of the gas plant from carburetted water gas to oil gas
production could have created variations in level or type of
pollution.
The fact that the excess lung cancer mortality has occurred
only in men raises the question as to whether an environmental
factor or differences in smoking characteristics or occupation
have caused the increased death rate in tract 2303. It is
possible that smoking and an environmental pollutant are
required to produce the excess of cancer. The rates in
older women then could be lower because they did not smoke
and the possible syngergistic effect of cigarettes and the
environmental factor were not observed. Additional
research to ascertain the rate of cigarette smoking in the
reported excess lung cancer is clearly necessary.
Supplementary Studies
In the meantime, further studies have been done to determine
whether the excess cancers might be associated with exposure
-------
to arsenic. Soil samples have been collected in the community
as a proxy measure of the presence of arsenic in the air at
a time when air sampling data were not available. Validation
of medical histories and the cell type of the cancers were
studied to determine whether these factors differed in
individuals whose lesions might have been caused by arsenic
because of residence in the index area. These factors .are
fully covered in the Appendices. However, let us look at
the soil sampling study and the validation of hospital
records and pathology of cancer cases in more detail.
Soil Sampling
As noted above, the purpose of the sampling of soil in the
area near the pesticide plant was to try to determine
whether arsenic had escaped into the community and whether
the soil level of this substance at varying distances from
the plant would suggest the direction of air spread of the
agent. A review of the experience of others in using a soil
level as a proxy measure of air values had demonstrated the
success of such a procedure. Landrigan et al. (11) had
tested lead and other metals including arsenic in an area
around a snelter. They had demonstrated that lead levels in
the air fell rapidly to a distance of 2-3 Km. from the plant
and more gradually thereafter until reaching background
levels at 4-5 Km. The dust sampling also indicated higher
levels closer to the plant except where there was shielding
-------
by mountains. Soil tests done in the same area indicated a
similar rapid dropoff in levels to 2-3 Km. but the values
remained elevated to longer distances of 10 Km. The tests
were run at the surface and at 2.5, 5.0 and 7.5 cm. depths
with the highest lead levels being consistently found at the
surface in this situation with continuous contamination.
The environmental levels correlated well with blood lead
levels in children. Milham and Strong (12) tested arsenic
levels in urine of children living near a smelter and found
a relationship with household dust. Dorn et al. (13)
studied the dustfall, soil, and vegetative levels of lead
and other metals originating from a lead smelter and correlated
these with levels in blood and milk of nearby cattle. The
results indicated that there were differences in the suspended
particulate level for metals by season with the winter
having the highest and the spring the lowest level and that
these levels correlated well with dustfall, with cattle body
burdens, and with the periods of greatest wind velocity.
Using these studies as models we presumed that arsenic might
act in a similar manner to lead and that soil levels could
represent previous air-borne exposures even though the
chemical plant did not have the tall stacks usually associated
with smelters and the surrounding geographic characteristics
of a city might alter the relationships reported from a
relatively open and rural type of environment.
The validity of our assumptions were investigated by contacting
-------
local consultants concerning the sampling of soil for
arsenic, the characteristics of Maryland soils, the possible
changes in arsenic with weathering in soil, and the influences
of local topography on deposition of arsenic. We- consulted
individuals from the U.S. Geological Survey Unit and the
Environment and Geography Department at the Homewood.Campus
as well as those from the Environmental Health Department at
the Medical Campus of the Johns Hopkins University. Individuals
from the University of Maryland and the Maryland State!
Department of Agriculture also provided information and
references. The consensus of opinion was that few tests for'
arsenic had been done in Maryland but it was felt that,
unless deliberate additions of arsenic-containing materials
.had occurred, very little of the metal should be present in
the soil naturally. Dr. Cady and others warned about
collecting samples under trees and in areas where wash-off
could occur since these situations could either reduce the
amount of dustfall from air-borne particles or dilute the
level of penetration into the soil.
The method of .selecting samples is covered in Appendix.I.
Results
The selection sites are shown in Figure 9.and the levels of
arsenic in Table 15. If we examine only the 101 samples
taken in the first set .we can see that, in general, arsenic
-------
levels were highest where lung cancer mortality was also
highest. The mean arsenic level from 20 sample sites in
tract 2303 was 63 ppm of arsenic. Even the omission of
samples from the park adjacent to the plant only reduced the
mean arsenic level to 38 ppm. Tracts 2301 and 2404 had
means of 6 ppm and 2302 a mean of 4 ppm based on only 2 to 4
sample sites. At most sites, the arsenic level was highest
at two inches. This suggested that recent contamination had
been less than in the psst. The highest and mean levels at
various sites are shown in the maps, Figures 10 and 11. If
we examine the samples omitting, sites 133, which were taken
during the 1977 spring period, we find the highest levels in
the park at the northern fence in an area which was adjacent
to the property line of the plant. It was also in this area
that one inch deep samples sometimes had the highest value
suggesting a more recent dustfall containing arsenic. All
sites in the park had high levels except for an area which
had been turned over and resodded and in which low arsenic
levels were present. The one inch levels near the fence
were as high as 695 and 226 ppm whereas at the opposite side
of the park the values were only 29 to 97 ppm at one inch
but as high as 46 to 161 ppm at two inches deep. From the
soil levels of the original samples, the data indicated high
levels near the plant to the east and north and south in the
park within a 3/8 mile radius of the plant. It was also
apparent that higher levels were found in a northerly
direction along the railroad lines and these levels had not
-------
dropped as far north as was tested in the first set of
samples. It was decided to collect additional samples to
the north along the railroad lines to determine the point at
which arsenic levels dropped to control values. Samples were
also collected to the east in the Port Covington yards and
along a spur to indicate levels associated with the presence
of coal and railroads.
The second set of samples include sites 1-33 listed as
Spring 1977. These samples, in general, had moderately high
levels of arsenic at three sites and there was no pattern of
a gradient in the levels at the sites most distant from the
plant. The level in the railroad yard sample was high with
a mean of 52.3 ppm. This value was higher than the other
sites east of the plant. Two sites to the north had unusually
high levels with means of 59.7 and 92.7 and highest values
of 144 and 173. These values do not follow a pattern of
decreasing levels at further distance from the plant.
Discussion
The initial samples indicated that arsenic had spread into
the general environment especially into the park adjacent to
the plant. The gradient of level by depth at each site
suggested that recent contamination had been less than
previously experienced pollution since the highest value
most frequently occurred at two inches and not near the
-------
surface. The levels at four inches were usually lower than
at the shallower depths but still correlated well with the
values at the surface. We felt the values at two inches
suggested generally long-term contamination of the area
although this could not be proven since we found no information
on the time required for percolation of arsenic through
soil. There was a gradient both to the north and the east
in earlier samples which suggested that wind currents toward
the east might have carried the material about 1/4 mile but
the spread was further in a northerly direction. It was
also apparent that the heaviest contamination was in the
park near the plant. It was impossible to detect high levels
in soil directly adjacent to areas where lung cancers
occurred although the higher levels of arsenic in a generally
northerly pattern correlated with the location of the
highest frequency of lung cancers as seen on the spot map.
It was proposed that the pattern of high levels might be
related to rail tansport of the material since an active
rail line did connect to the company. Fugitive emissions
from this transportation route might have accounted for the
erratic distribution of arsenic by gradient of level.
The further samples collected at sites in northerly and
easterly directions along railroad lines and into the
railroad yards created some questions about the spread of
arsenic from the plant. The levels in the Port Covington
yard were high with values of 41 to 60 ppm arsenic at a site
-------
approximately 10 feet from the railroad tracks in a flat
area with no run-off. At first we thought these values
could be explained by the presence of coal dust in the soil.
Mr. Janus, chief of coal sampling and inspection from the
Bureau of Mines provided information on the components of
the Pittsburgh and Upper Freeport seam coals which are the
types found in the Baltimore years. As reported by Kesslerr
Sharkey and Friedel (14), neither the ash nor the dust has
more than 16 ppm of arsenic, a value much lower than those
observed.
The second possible explanation for the presence of arsenic
in the yards was that pesticides were transported there in
open cars. That would not be possible under current conditions
since the yard handles primarily containerized freight
except for coal. There are no records of what materials it
handled previously and how they were transported. It seems
unlikely that arsenic in open cars would have contaminated
the yards without also leaving material along the tracks
nearer the plant. No sites were sampled along those tracks
and, thus, further sampling in these areas is needed.
The third possibility was that herbicides had been used
around the yards and along railroad lines to control vegetation,
There are no records of any herbicide use in the area. The
high levels at all depths would suggest repeated high
concentrations over long periods as one would expect with
-------
herbicide treatments. However, the levels along a smaller
track to the north shows no signs of high levels of this
magnitude. If herbicides are the explanation there must
have been selective use of the product on different railroad
beds.
High levels are seen north of the natural gas plant and
there is no gradient by distance from the plant as expected.
The sites sampled in these areas-were at least 30 feet from
the railroad track, a distance which is probably too far to
be contaminated by herbicides although there could be
distant run-off from applications to the bed. We could
postulate that material had been moved from the plant in a
northerly direction with varying degrees of contamination of
soil depending on track slope or other factors but further
information is needed especially to explain the lack of a
gradient.
Additional consultation was sought from Dr. Woolson at the
Pesticide Degradation Laboratory in Beltsville. Except for
the suggestion that sodium arsenite or perhaps the methanearsonates
used on the railroad beds might play a role, he had no other
recommendations. Additional information could be sought
from the company about methods of transportation if further
sampling is anticipated so that more appropriate sites can
be selected for study.
-------
There are many other possible steps which could be taken.
Information is needed on the time of contamination which
would be represented by the presence of arsenic at a depth
of 2 inches. This might require simulated conditions if
there are not data from other sources. We should also
examine the possibility of testing the sediment in the
surrounding waters for arsenic unless pollution from multiple
sources represents too much of a problem. Sites from the
second sampling set could be repeated using closely adjacent
spots to see if the discrepant levels represented some type
of local contamination. Many of the areas in that vicinity
have had recent changes in topography, however, due to
construction of a bridge in that neighborhood which may have
eliminated, some of the original sites.
Validationof hospital records and cancer pathology
This phase of the study of the possible association of
mortality with exposure to pesticides included a review of
hospital records of patients. (Appendix II) The purposes of
this phase were to determine the accuracy of death certification
of cancers in Baltimore; to identify any possible differences
in diagnosis by area; to determine any variation in the
pathological characteristics of cancers especially those of
the lung in the index census region as compared to others;
and to investigate any possible differences in personal
characteristics as described in hospital records. The
-------
investigation would focus specifically on possible identification
of unusual cell types of lung cancer in the presumed
arsenic-exposed group and the presence of histories of
known arsenic-associated diseases in the patients dying of
lung cancer in the index area.
Results
The causes of death were grouped into two time periods which
represented the use of the 7th and 8th ICDA codes and
grouped into causes as listed on the certificate by the
first two digits of the code. These causes were then
compared to the first four medical conditions or diagnoses
as noted on the hospital records. The individuals in whom
the records did correspond within the first two digits for
lung cancer and total cancer only are noted on Table 16. It
is apparent that 15.0 to 18.5 percent of cancers identified
by death certificate do not have the same diagnosis as
listed among the first four diagnoses on hospital records.
The problem of validation of death certificate information
was reviewed further by a physician who examined the data on
the abstract forms. As indicated in Table 7, there was
complete agreement in diagnoses to four digits in the ICDA
code in only 75.0 percent of the cancer deaths. If classification
to three digits only is used we will correctly verify 80.7
percent of the cases listed on the certificate. In 1.8
-------
percent of case metastatic lesions were identified on the
death certificate as underlying and in another 5.5 percent
multiple cancers were listed on the certificate and the
primary site varied from that listed on the hospital record.
There was no cancer diagnosis listed anywhere on the hospital
record for 2.7 percent of deaths. A further examination of
the method of diagnosis of cases was attempted in order to
demonstrate whether differences in the methods of diagnosis
might have changed the accuracy of death certification.
Data from autopsy or histological examination of tissue
was used for the diagnosis for 82.9 percent of the cases
with complete agreement in records and 88.9 percent of cases
where the agreement was less than perfect. Therefore, the
reliability of the hospital diagnosis as determined from the
method by which the cancer was identified did not select
those cases which were inappropriately classified on the
death certificate.
The method by which the diagnosis of cancer was made differed
only slightly in the larger hospitals with 69.7 to 91.4
percent of cases diagnosed by autopsy or histology. An
examination of the differences in diagnosis by census tract
has not been completed, but it is unlikely that there will
be variations in diagnosis since we have included all
hospitals used by individuals from the index area in the
above evaluation. An examination of the hospital records
for possible arsenic-associated symptoms included
-------
gastrointestinal signs, skin lesions, Mee's lines on nails,
neurological or neuromuscular symptoms, cardiovascular
disease, stroke and asthma. Only respiratory symptoms were
slightly higher in the index tract but since there were so
many records in which there was no comment about these
symptoms, it is difficult to interpret the small variation.
We also sought information on diseases for which arsenic
might have been used as a treatment, such as syphilis,
trypanosomiasis and amebiasis and the results indicated no
higher frequency of these conditions among residents of
index tracts. As might be expected, most of these diseases
were rarely recorded for deaths in any area.
Both smoking and drinking histories were abstracted from
hospital records. Drinking was rarely recorded and smoking
histories were frequently missing. Table 18 indicates the
smoking characteristics of lung cancer deaths in index and
control tracts as determined from the hospital record. For
46.6 percent of the patients, the smoking histories are
unknown. Despite that fact, we have attempted to compare
the smoking levels in index versus the control tracts. The
percent of smokers is slightly higher in the index tracts
but the difference does not appear impressive. If one
includes only those charts with a recorded history, almost
all cases are positive for smoking in both index and control
tracts.
-------
The original hypothesis was that if arsenic had caused the
lung cancers, the cell type of lesions from the index tract
might differ compared to other areas with an expected
predominance of small cell or oat cell tumors in the exposed '
tracts. The data in Table 19 would indicate that the cell
types differ very little from index to control tracts. '
Discussion
Despite the suspicion that the characteristics of individuals
in tract 2303 might differ from control tracts because of
exposure to arsenic, there were no indications from the data
that they had an increased risk of arsenic-associated
diseases. They also did not have differences in smoking
histories or in the frequency of the cell types of lung
cancers. If we could have shown an excess risk of smoking
in tract 2303, we might have explained the increased mortality
from lung cancer on the basis of differences in cigarette
consumption. If we could have shown an increased frequency
in specific cell types among the lung cancers from tract
2303 as compared to controls, we could have suggested that
exposure to a pollutant other than cigarette smoke was the
probable cause of the increased lung cancer mortality.
Neither situation occurred. Actually, the number-of charts
which had adequate records to answer'questions about smoking
were very few. Only a survey would provide the data which
are needed for such a review.
-------
It is apparent from the data that the classification of
deaths from death certificates are often in error compared
to the original hospital record. However, if we adhere to
major categories of cancer, most of the records do correspond.
For about three percent of the records, there is no clue as
to how the diagnoses were determined. It is possible that
these were the wrong records despite the multiple methods of
identification. It is also possible that a history of
cancer was missed in the abstracting. It is safe to say
that death certificate information for Baltimore City is
usually accurate in identifying a cancer if the first two
digits of the classification only are considered. Therefore,
the excess risk of lung cancer identified in tract 2303 is a
real increase in mortality and not the result of local
diagnostic and certification practices.
-------
Conclusion:
The data from a study of cancer mortality in an industralized
area of Baltimore where an arsenical-producing pesticide
plant was located indicated a significant excess risk of
death from lung cancer and all cancers among males. The
risk has increased in the period around the 1970 census as
compared to previous time intervals. No risk of lung cancer
was found in females nor in residents of census tracts
located over one-half mile from the plant. The excess was
not due to occupational exposure to pesticide manufacture.
Soil levels of arsenic were higher in the areas where
the mortality from lung cancer was also higher. The arsenic
levels appeared to be increased in the area around the
north-bound railroad line. Although there was a gradient of
reduced arsenic levels with increased distance from the
plant, it was not a perfect correlation for all sites along
the railroad, possibly because of erratic fugitive emissions
from a transport route.
The excess lung cancer mortality in the local area
could not be attributed to differences in diagnostic practices
between hospitals in the city. Smoking histories in hospital
records were very incomplete but did not indicate differences
in cigarette use in index and control census tracts. No
symptoms of arsenic-related disease could be identified on
the hospital records. The frequencies of specific cell
types among the lung cancers in the index tracts and control
tracts were similar.
-------
Further research is required to determine whether
smoking patterns, non-pesticide associated occupational
exposure or residential history may play a role in the
excess of lung cancer.
-------
References
1. Tseng, W. P., H. M. Chu, S. W. How, J. M. Fong,
C. S.. Lin, and S. Yeh, Prevalence of skin cancer in
an endemic area of chronic arsenicisra in Taiwan. J.
Nat. Cancer Instit. 40: 453-463, 1968. -
2. Braun, W. Carcinoma of the skin and the internal
organs caused by arsenic: Delayed occupational lesions
due to arsenic. German Med. Monthly 3: 321-324, 1958.
3. Neubauer, 0. Arsenical cancer: A review. Brit. J.
Cancer 1: 192-251, 1947.
4. Ott, M. G., B. B. Holder, and H. L. Gordon, Respiratory
cancer and occupational exposure to arsenicals.
Arch. Environ. Health 29: 250-255, 1974.
5. Baetjer, A., A. Lilienfeld, and M. Levin, Cancer and
occupational exposure to inorganic arsenic, p. 393.
In Abstracts. 18th International Congress on
Occupational Health, Brighton, England 14-19, Sept.
1975.
6. Mabuchi, K., A. M. Lilienfeld and L. M. Snell, Lung
cancer among pesticide workers exposed to inorganic
arsenicals. Arc. Env. Hlth. 312-320, Sept/Oct, 1979.
7. Blot, W. and J. Fraumeni, Arsenical air pollution and
lung cancer. Lancet, (7926): 142-144, July 26, 1975.
8. Matanoski, G. M., E. Landau and E. Elliott, Pilot
study of cancer mortality near a chemical plant and
other industries in Baltimore. U.S. Environmental
Protection Agency, Office of Toxic Substances, EPA
560/6-76-003, Final Report, May 1976.
9. B. Woolf, "On estimating the relation between blood
group and disease," Ann Eugen 19: 251 (1955).
10. J. B. S. Haldane, "The estimation and significance
of the logarithm of a ratio of frequencies,"
Ann Hunan Genetics 20: 309 (1956).
11. Landrigan, P. J., S. H. Gehlbach, B. F. Rosenblum
J. M. Shoults, R. M. Candelaria, W. F. Barthel, J.
A. Liddle, A. L. Smrek, and N. W. Staehling, Epidemic
lead absorption near an ore smelter; The role of
particulate lead. New Eng. Jour. Med. 292:
123-129, 1975.
12. Milham, Jr. S. and T. Strong, Human arsenic exposure
in relation to a copper smelter. Environ. Res. 7:
176-182, 1974.
-------
13. Dorn, C. R., J. O. Pierce, G. R. Chase and P. E.
Phillips, Environmental contamination by lead,
cadmium, zinc, and copper in a new lead-producing
area. Environ. Res. 9: 159-172, 1975.
14. Kessler, T., A. G. Sharkey, Jr., and R. A. Friedel,
Spark-source mass spectrometer investigation of coal
particles and coal ash. Bureau of Mines Respirable
Dust Research Program, Technical Progress Report-42,
Pittsburgh Energy Research Center, Pittsburgh, Pa.,
1971, (U.S. Department of the Interior).
-------
Table 1.
Designation of 1970 Census Tracts
Match I
Match II
* Plant
Index
Tracts
2302
2303*
2404
2301
located in
Adjacents
Controls
2401
2402
2403
2101
2201
this tract.
Table 2
South
Control
2005
2502.05
2503.03
2504.01
2504.02
2505
Designation of 1960 Census
Match I
Match II
Index
Tracts
2302
2303
2404
2301
Adjacent South
Controls Controls
2401
2402
2403
2201
1903
1904
2003
2005
2008
2504"
2505
2503B
2502C
Central
s Controls
101
104
105
601
602
2608
703
Tracts
Central
Controls
101 703
102 2603
103 2607
104 2608
105 2609
201 2610
202 2611
203 2604A
601 2606A
602 2604B
701
702
604
North
Controls
1308.02
2604.01
2702
905
1203
-
North
Controls
904
905
906
1305
1306
1308
1512
1513
2716
2718
2801A
1104
1205
1401
-------
Table 3. Characteristics of Index and Control Census Tracts from 1970 Census for Selected Baltimore Areas
Characteristics
Total population
Range between tracts
X < 5 yi s
£ b-19 yrs
% ?0-44 yrs
Z 45-64 yrs
X 6tHyrs
X Male
% Elar.k
Median Income
(Family)
Match
2303
1703
B
29
26
28
9
50
-
8400
I
2302
3395
B
28
20
24
11
50
0
9200
2404
3697
8
28
29
25
10
48
-
8400
Adjacent
Controls
9347
2872r3358
6-8
25-26
28-29
23-38
10-16
4B-49
0-2
8100-9300
South
Controls
31,174
3143-6893
6-12
24-30
27-39
14-28
4-13
47-49
0-2
7500-9000
Central
Controls
20.37B
2125-4055
6-8
23-28
26-32
25-33
12-15
46-50
0-2
7500-9300
North
Controls
12.964
2316-5969
6-10
23-26
26-35
22-27
6-17
47-49
0-4
8800-10000
Match
2301
3560
10
30
26
23
10
49
50
6500
II
Adjacent
Controls
5606
1935-3671
a
28-30
25-26
25-26
12-13
48-54
50-52
5800-6000
Central
Controls
3180
10
30
26
23
12
40
47
6800
North
Controls
8139
2952-51B7
9
29-33
32-34
17-19
8-10
47-48
39-64
7000-7300
-------
Table 4. Characteristics of Index and Control Census Tracts from 19GO Census for Selected Baltimore Areas
aractertstics
tal population
nge between tracts
«5 yrs
5-19 yrs
20-14 yrs
C5-64 yrs
65+yrs
I-'ale
Black
dian Incon*
(Family)
Match
2303
2135
10
29
31
22
a
51
1
5600
I
2302
3924
10
26
31
22
10
48
1
5100
2404
4204
10
27
31
22
9
50
0
5600
Adjacent
Controls
10.529
3178-3682
10-11
27
31-33
21-24
9
49-50
0-1
5500-6200
South
Controls
49,997
2562-12163
8-12
21-28
28-37
17-27
5-16
47-52
0-8
4800-6200
Central
Controls
1 03.839
2548-17511
7-15
20-33
23-38
15-31
2-14
47-53
0-13
4800-6300
North
Controls
53.874
1762-6264
6-11
17-27
25-35
21-29
8-22
44-49
1-12
4900-6300
Hatch II
2301
4100
13
28
30
22
9
49
60
3800
Adjacent
Controls
3253
11
24
32
25
8
55
49
3600
Central
Controls
4367
16
37
25
17
6
50
51
-3300
North
Controls
15.133
3134-6043
12-14
20-23
33-39
18-20
8-12
46-51
47-71
3300-4100
-------
Table 5. Duration of Residence in Household
1960*
Match I
Index
Adjacent Controls
South Controls
Central Controls
North Controls
Match II
Index
Adjacent Controls
Central Controls
North Controls
Mean
58
62
54
57
56
46
44
34
23
Range
54-65
55-70
43-65
18-68
49-61
46
44
34
22-32
1970*
Mean
64
67
62
63
64
59
60
49
48
Range
55-69
55-76
57-72
50-69
60-67
59
54-67
49
44-53
* 1960 % living in same house-7+ years
1970 % living in same 5+ years
-------
Table 6
Cause
Average Annual Crude Death Rate per 100,000 for Index and Control Census Tracts
Deaths 1958-62
Hales
Match I
2303 2302 2404 Adjacent South Central North
# Rate S Rate § Rate « Rate if Rate # Rate f Rate
Oral
Pancreas
Lung
Prostate
All Cancer
Cause
Oral
Pancreas
Lung
Prostate
All Cancer
0
1
4
2
11
if
0
0
3
1
10
-
18.5
74.1
37.1
203.9
2301
White
Rate
-
-
75.2
25.1
250.6
2 21.1
0
12 126.8
2 21.1
26 274.8
Nonwhite
# Rate
0
1 16.8
2 33.5
2 33.5
16 268.2
1 9
2 19
8 75
3 28
32 303
.5 5
.0 3
.9 23
.4 3
.5 56
Match I
Adjacent
White Nonwhite
*
1
0
2
1
Rate
21.8
-
43.6
21.8
8 174.3
# Rate
0
1 22.9
1 22.9
1 22.9
12 274.9
19.3 10
11.6 8
88.8 79
11.6 19
216.1 244
I
8
6
65
15
202
.3 20
.6 31
.4 193
.7 43
.1 615
Central
White Nonwhite
# Rate
0
1 17.7
5 88.7
0
14 248.2
|
0
1
6
0
17
Rate
-
18.9
113.6
-
322.0
7.6 11
11.8 18
73.6 60
16.4 23
234.6 293
8.9
14.6
48.8
18.7
238.2
North
White Nonwhite
# Rate #
3 10.4 0
2 12.2 0
12 73.4 8
5 30.6 3
43 263.2 39
Rate
-
-
39.3
14.7
191.6
-------
Table 7
Average Annual Crude Death Rate per 100,000 for Index and Control Census Tracts
Deaths 1968-74
Males
Match I
Census Tracts
Cause
Oral
Pancreas
Lung
Prostate
All Cancer
Cause
Oral
Pancreas
Lung
Prostate
All Cancer
2303
2302
H Rate 0
2 33
2 33
18 305
1 16
33 559
2301
White
9 Rate
0
0
7 113.3
1 16.2
17 275.0
.9
.9
.0
.9
.2
0
0
12
1
26
Rate
-
-
103.3
8.6
223.8
2404 Adjacent
tt
0
1
12
2
24
Rate J
4
8.0 3
96 . 1 24
16.0 4
192.2 86
Match II
Rate
12.7
9.6
76.4
12.7
273.9
Adjacent
Bl
#
1
3
13
3
27
ack
Rate
16.9
50.7
219.5
50.7
455.9
Wh1
#
2
0
13
4
34
te
Rate
20.5
-
133.1
41.0
348.2
Black
if Rate
2 20.3
3 30.5
16 162.6
4 40.6
42 426.7
South
#
7
9
111
IS
254
Rate
6.8
8.8
108.3
14.6
247.9
Central
tt
6
9
60
25
206
Rate
9.0
13.4
89.6
37.3
307.7
North
ft
5
3
48
7
110
Central
White
' 0
1
2
4
1
17
Rate
17.5
35.1
70.1
17.5
290.0
Black
#
0
0
5
2
13
Rate
-
-
102.8
41.1
267.2
White
ti
0
0
15
3
41
Rate
11.7
7.0
112.4
16.4
257.7
North
Black
Rate .# Rate
-
-
124
24
340
1 6.9
1 6.9
.5 6 41.3
.9 3 20.7
.3 1 28 192.9
-------
Table 7A
Crude Lung and All Cancer Deaths Rates per 100,000
White Males in Census Tracts the Same in 1960 and 1970
2303
Census Tracts
2302 2404
Adjacent Central
Lung Cancer
1958-62
1968-74
All Cancer
1958-62
1968-74
74.1
305.0
203.9
559.2
126.8
103.3
274.8
223.8
75.9
96,1
303.5
192.2
88.8
76.4
216.1
273.9
77.7
39.6
241.5
307.7
-------
Table 8
Average Annual Crude Death Rate per 100,000 for Index and Control Census Tracts
Deaths 1958-62
Females
Hatch I
Cause
Oral
Pancreas
Lung
Breast
Cervic
Al 1 Cancer
2303
it Rate
0
0
0
3 57.5
0
6 115.1
2302
* Rate
0
2 19.9
0
2 19.9
1 10.0
17 169.2
2404
# Rate
0
0
0
3 28.7
1 9.6
13 124.4
Adjacent South
# Rate #
0
0
6 22
8 30
3 11
43 162
2
8
.6 8
.2 28
.3 15
.3 186
Rate
1.6
6.4
6.4
22.5
12.0
149.3
Central
ti
2
23
18
81
34
460
Rate
0.7
8.4
6.5
29.4
12.4
167.2
North
ft
4
12
15
51
12
263
Rate
2.9
0.8
11.0
37.5
8.8
193.3
Match II
Cause
Oral
Pancreas
Lung
Breast
Cervix
All Cancer
2301
White
# Rate
0
0
0
3 71.4
1 23.8
G 142.9
Nonwhite
9 Rate
0
0
1 15.8
2 31.5
1 15.8
10 157.6
Adjacent
White
# Rate
1 27.4
1 27.4
0
3 82.2
1 27.4
10 274.0
Central
Nonwhite
tt Rate
0
1 27.3
0
1 27.3
1 27.3
8 218.6
White
9 Rate
0
0
1 19.7
2 39.5
2 39.5
9 177.7
Nonwhite
tt
0
0
0
1
1
6
Rate
-
-
-
17.1
17.1
102.6
North
White
H
0
1
1
8
5
41
Rate
-
5.8
5.8
46.6
29.1
238.9
Nonwhite
9 Rate
0
0
0
4 18.
3 13.
21 96.
3
8
3
-------
Table 9
Average Annual Crude Death Rate per 100,000 for Index and Control Census Tracts
Deaths 1968-74
Females
Match I
Cause
Oral
Pancreas
Lung
Breast
Cervix
All Cancer
Cause
Oral
Pancreas
Lung
Breast
Cervix
All Cancer
#
2
0
1
1
0
9
*
0
1
2
5
0
15
2303
Rate
33.3
-
16.6
16.6
-
149.9
2301
White
Rate
-
16.4
32.8
82.0
-
246.0
2302
2404
1 Rate t
0
2 16
3 25
1 8
3 25
14 117
0
.8 2
.2 1
.4 2
.2 0
.8 17 1
Adjacent South
Rate
-
15.0
7.5
15.0
-
27.3
8
4
. 1
7
8
3
56
Match II
Adjacent
Nonwhlte
# Rate
0
0
1 15.4
3 46.2
2 30.8
18 277.1
White
1 Rate
0
1 10.9
2 21.9
6 65.7
3 32.8
26 284.6
Nonwhlte
»
1
2
7
4
1
25
Rate
9-7
19.3
67.6
38.6
9.7
241.5
Rate #
12.1 3
3.0 11
21.1 16
24.1 25
9.1 6
169.0 165
Rate
2.7
9.9
14.4
22.5
5.4
148.7 1
Central
White
# Rate
0
1 16.7
1 16.7
2 33.4
0
14 233.6
Nonwhlte
# Rate
0
0
1 18.1
0
1 1B.1
5 90.5
Central
0 Rate
3 4.1
10 13.6
18 24.5
29 39.5
5 6.8
75 238.2
North
White
# Rate
0
3 23.0
3 23.0
8 61.4
2 15.4
34 261.1
North
#
2
1
9
9
0
78
Rate
4.3
2.2
19.5
19.5
-
169.2
Non white
f
0
0
0
3
1
16
Rate
-
-
-
18.1
6.0
96.3
-------
Table 10
Average Annual Crude Rates per 100,000
Male Lung Cancer
Match I
Years 2303 2302 2404 Adjacent South Central North
1953-62
1966-67
1968-72
1973-74
74
177
284
355
.1*
.9*
.7*
.9*
126.8
180.7
120.5
60.2
75
84
100
84
.9
.1
.9
.1
83.
in.
66.
100.
8
5
9
3
65.4
54.6
no. 7
102.5
73
151
85
99
.6
.6
.7
.3
48.8
82.0
114.8
106.6
Significance tests * P^-0^: Compared to Northern Controls.
Woolf-Haldane relative risk analysis.
Match II
2301 Adjacent Central North
Years White Black* White Black White Black White Black
1953-62
1966-67
1968-72
1973-74
75.2
56.6
90.6
169.9
33.5
-
260.0
178.2
43.6
358.4
129.0
143.4
22
248
184
106
.9
.9
.9
.7
88
61
98
-
.7
.3
.2
132.6
71.9
115.1
71.9
73
145
93
203
.4
.3
.0
.4
39.3
-
38.6
48.2
* 1958-62 rates are for ncnwhites.
-------
Tin»e-Adjusted SMRs Based on Average Annual Baltimore City Rates*
Deaths 1958-1962 and 1966-1974
Match I White Males
Cause
Oral
Stomach
Colon
Rectum
Pancreas
Lung
Prostate
Bladder
Lymphomas
All Cancer
Cause
Oral
Stomach
Colon
Rectum
Pancreas
Lung
Breast
Cervix
Bladder
Lymphomas
All Cancer
obs
3
2
3
1
5
25
3
0
2
54
obs
2
1
1
4
0
1
5
0
0
1
18
2303
SMR
2.45
1.60
1.22
1.09
4.15
2.74
1.59
-
0.88
1.94
2303
SMR
6.31
1.32
0.45
6.12
-
0.71
1.29
-
-
0.65
0.96
2302
obs
2
3
9
3
0
30
3
3
4
67
SMR
0.88
1.18
1-74
1.62
-
1.72
0.68
1.3R
0.89
1.21
2302
obs
0
3
7
0
4
4
5
5
0
4
41
SMR
-
1.74
1.36
-
2.32
1.34
0.61
2.30
-
1.19
1.00
2404
obs
1
4
6
4
3
23
5
2
2
63
SMR
0.43
1.59
1.21
2.19
1.31
1.31
1.26
0.97
0.44
1.15
2404
obs
1
0
5
2
2
1
6
3
0
3
39
SMR
1.40
-
0.94
1.32
1.13
0.31
0.68
1.27
-
0.86
0.91
Adjacent
obs
9
10
9
6
7
57
7
5
10
168
Match I
Adjacent
obs
4
5
13
3
1
15
17
6
0
6
109
Controls South
SMR
1.43
1.44
0.62
1.18
1.09
1.19
0.56
0.85
0.80
1.10
White
obs
18
23
43
14
21
206
36
20
26
543
Females
Controls South
SMR
2. 06
0.96
0.83
0.69
0.19
1.70
0.70
0.97
-
0.61
0.90
obs
5
14
41
14
20
26
50
25
10
36
390
Controls
SMR
0.88
0.98
0.94
0.85
1.01
1.35
0.93
1.03
0.62
1.10
Controls
SMR
0.84
0.89
0.89
1.08
1.34
0.99
0.75
1.13
1.33
1.22
1.05
Central
obs
29
60
67
34
43
282
74
37
65
891
Central
obs
5
37
81
33
30
37 .
119
43
15
52
682
Control s
SMR
0.91
1.40
0.91
1.25
1.27
1.21
1.15
1.11
1.02
1.13
Controls
SMR
0.57
1.30
1.01
1.57
1.59
1.05
0.94
1.14
1.15
1.05
1.10
North
obs
16
17
38
15
2\
118
31
28
37
423
North
obs
6
18
40
13
13
23
61
13
6
24
350
Controls
SMR
0.09
0.70
0.90
0.97
1.10
0.93
0.79
1.45
1.06
0.96
Controls
SMR
1.14
1.01
0.82
1.03
0.90
1.14
0.82
0.61
0.74
0.83
0.96
* AvcL'ayc annual Baltimore City rates (based on deaths In 1950-1962) were applied to the 1960 match population and
were weighted for five years. Averaye annual Baltimore City rates (based on deaths In 1960-1972) were applied to the
IT'fl mil h popular inn -,,ri(\ \.tr>r<> w-iqlifpr| fnr nino voars.
-------
Table 12
Age-and Time-Adjusted Mortality Rates per 100,000 for Lung Cancer
Males
Match I
Years
1958-59
1960-62
1966-67
1968-69
1970-72
1973-74
2303
IV
0
4
3
3
9
6
Rate
-
166.8
330.1
180.0
349.7
512.5
2302
8
7
5
6
3
7
2
Rate
216.1
107.5
206.7
104.4
205.7
61.7
Z404
#
2
6
3
7
2
3
Rate
57.8
119.2
123.8
241.2
38.6
106.1
Adjacent
9
11
12
10
5
10
9
Rate
141.4
99.7
119.9
61.8
85.1
104.9
South
Central North
# Rate 0
35 98
44 84
16 76
30 141
51 155
30 139
.2 79
.9 114
.1 29
.4 10
.9 31
.3 19
Rate V
94.8 22
89.9 38
153.5 10
52.7 16
109.3 19
100.0 13
Rate
49.1
55.3
93.6
166.4
124.7
126.3
Match II
Year
1958-59
1960-62
1966-67
1968-69
1U70-72
1973-74
White
2301
NW
0 Rate #
2 140
1 55
1 61
0
4 181
3 271
.8 0
.1 2
.0 0
4
.7 7
.4 2
Adjacent
or B*
Rate
-
51.2
-
141.7
191.7
62.5
White
# Rate
1 46
1 38
10 322
3 97
6 122
4 121
.7
.5
.0
.7
.2
.2
NW
f
0
1
7
9
4
3
or B
Rate
Wh1
#
Central
te
Rate
4 223.0
17.7
160.4
173.7
65.7
62.0
1
1
1
3
0
36.6
35.1
44.5
81 .0
_
NW or B
# Rate
3 144.6
4 136.5
1 87.1
1 98.5
3 258.8
1 196.3
North
White NU
# Rate #
5 83.0 3
7 83.4 5
5 149.0 0
5 155.4 1
3 57.4 3
7 204.5 2
or B
Rate
42.4
45.2
-
32.0
67.2
66.8
1958-59 and 1960-62 are non-white
Al1 others are black.
-------
Table 12A
Age-and Time-Adjusted Mortality Rates per 100.000 for Lung Cancer
Employee deaths removed
Males
Match I
Years
1958-59
1960-62
1966-67
1968-69
1970-72
1973-74
Year
1958-59
1960-62
1966-67
1968-69
1970-72
1973-74
ft
0
4
3
2
a
6
White
2303
Rate
-
166.8
330.1
94.6
309.5
512.5
2301
NW
i Rate t
2 140
1 55
1 61
0
4 181
3 271
.8 0
.1 2
.0 0
4
.7 7
.4 2
2302
It
7
5
5
3
7
2
Rate
216.1
107.5
169.5
104.4
205.7
61.7
2404 Adjacent
I
2
5
3
7
2
3
Rate |
57.8 11
104.1 12
123.8 10
241.2 5
38.6 9
106.1 9
Match II
Rate
141.4
99.7
119.9
61.8
78.4
104.9
Adjacent
or B*
Rate
-
51.2
-
141.7
191.7
62.5
White
it
1
1
10
3
6
3
Rate
46.7
38.5
322.0
97.7
122.2
88.4
NW or B
# Rate
0
1 17.7
7 160.4
9 173.7
4 65.7
3 62.0
South Central North
*
35
44
16
30
51
30
Rate #
98.2 79
84.9 114
76.1 29
141.4 10
155.9 31
139.3 19
Central
White
//
4
1
1
1
3
0
Rate
223.0
36.6
35.1
44.5
81.0
.-
NW or B
# Rate
3 144.6
4 136.5
1 87.1
1 98.5
3 258.8
1 196.3
Rate #
94.8 22
89.9 38
153.5 10
52.7 16
109.3 19
100.0 13
North
White
# Rate
5 03.0
7 83.4
5 149.0
5 155.4
3 57.4
7 204.5
Rate
49.1
55.3
93.6
166.4
124.7
126.3
NW or B
if Rate
3 42.4
5 45.2
0
1 32.0
3 67.2
2 66.8
* 1958-59 and 1960-62 are non-white
All others are black.
-------
Table 13
Age-and Time-Adjusted Mortality Rates per 100,000 for Lung Cancer
Females
Match 1
Years
1958-59
1960-62
1966-67
1968-69
1970-72
1973-74
2303
df Rate
0
0
0
1 123.5
0
0
2302
8 Rate
0
0
1 28.8
0
3 66.2
0
2404
.# Rate
0
0
0
0
0
1 27.3
Adjacent
# Rate
3
3
2
0
4
3
48.2
32.6
22.4
-
31.3
31.7
South
# Rate
2
6
2
6
4
6
4.6
15.2
11.4
21.9
11.2
23.0
Central
JH Rate
7
11
1
7
6
5
9.0
10.7
5.2
34.2
17.8
26.4
North
// Rate
9
6
0
i:
5
2
20.3
10.3
-
6.4
30.9
15.5
Match II
Years
1958-59
1960-62
1966-67
1960-69
1970-72
1973-74
2301 Adjacent
White NW or B* White NW or B
H Rate rff Rate # Rate f Rate
0
0
0
0
1
1
0
1
0
1
36.3 0
67.1 0
0 -
20.6 0
0 -
27.3 1 43.1
1 27.6
0
o -
0
0
2 45.3
2 22.5
3 69.2
White
ft Rate
1 142.9
0
0
1 61.6
0
0
Central
NW or B
# Rate
0
0
0
0
1 56.5
0
North
White NW or B
tt Rate # Rd
0
1
2
0-
1
2
-
11.0
50.6
-
16.5
51.3
o
0
0
0
0
0
* 1958-59 and 1960-62 are non-white
All others are black.
-------
Table 14
LIMITS ON SMR BASED ON A POISSOM DISTRIBUTION
Census Tract
2303*
0203*
2005*
2006
1205*
1207
2533
2553*
2511*
2636*
2653
2664
2737
2765
White Male Lung Cancer
Obsa Expb SMR
9
9
15
7
7
8
5
9
11
8
2
6
2
4
2
4
6
3
2
3
1
3
4
3
0
2
0
2
.14
.00
.10
.15
.52
.92
.72
.95
.15
.07
.98
.14
.98
.00
4.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
21
25
46
22
78
04
91
28
05
61
05
19
05
00
1
1
1
0
1
0
0
1
1
1
0
0
0
0
LL
.93
.03
.38
.89
1.2
.88
.94
.04
.32
.13
.25
.80
.25
.54
7
4
4
4
5
4
6
4
4
5
7
4
7
5
UL
.98
.27
.06
.58
.72
.02
.78
.33
.74
.13
.37
.77
.37
.12
* 95% confidence lim'te does not include
probability of 1.00 for the SMR
a Observed deaths in 1970-1972
b Expected deaths based on the average annual
Baltimore City rate. Deaths in 1970-1972.
-------
Table
Sampling Sites and Arsenic Content (ppm) by Depth of Soil
Sunv.r 1976
ppm As ppm As
Site
1-4
5-8
9-12
13-16
17-18
19-20
21
22
23-25
26-28
29-31
1"
97
29
226
87
14
6
8.9
8.9
76.9
96.5
65.1
32-34 163
35-37
38-40
41-43
44-46
47-49
50-52
41.0
63.8
34.6
63.7
25.7
6.1
2"
161
46
93
101
102
67.8
137
85.7
100
126
38.8
73.1
45.6
3.5
3" 4"
86 86
21 14
80 48
59 69
7
4
18.7
56.5
15.2
22.1
46.8
102
54.7
13.2
17.8
16.9
Average of
1", 2", 4"
114.7
29.7
122.3
85.7
65.9
73.6
72.4
90.3
62.6
97.3
42.7
50.0
29.7
8.8
Site
53-55
56-58.
59
60-62
63-65
66-68
69-71
72-74
75-77
78-80
81-83
84,86
87-89
90-92
93-95
96-98
99-101
-
1"
5.1
5.8
7.1.
4.0
1.5
695
9.0
8.9
8.9
3.2
11.5
2.6
3.0
5.0
4.5
6.2
11.0
2"
2.5
8.1
1.1
7.5
55.2
5.4
9.7
12.5
7.7
12.7
3.1
2.1
4.6
2.8
4.4
7.5
4
5
3
4
24
4
5
5
5
11
4
19
2
4
2
6,
4"
.4
.2
.0
.0
.6
..o
.3
.4
.2
.4
.6
.0
.8
.5
.8
.5
Average of
1", 2", 4"
4
6
2
4
258
6
8
8
5
11
3
8
4
3
4
.0
.4
.7
.3
.3
,1
.0
.9
.4
.9
.4
.0
.1
.9
.5
8.3
Spring 1977
Site
1-3
4-6
7-9
10-1
13-1
16-1
2
5
8
1" 2"
17 5
9 13
10 8
41 56
4 22
14 20
4"
6
4
2
60
19
12
Average of
1", 2", 4"
9.3
8.7
6.7
52.3
15.0
15.3
Site
19-21
22-24
25-27
28-30
Control
30-33
1"
12
6
4
90
5
2"
34
29
9
173
7
4"
12
144
20
15
4
Average of
1", 2", 4"
19.3
59.7
11.0
92.7
5.3
-------
Table 1 e
Non-Agreement of Certificate and Hospital Record of
Cancer for Lung and Total
Lung Cancer Total
1958-67
Total 99 414
Non-agreement 15 62
Percent 15.2 15.0
1968-74 .
Total 119 356
Non-agreement 20 66
Percent 16.8 18.5
-------
Table 17
Level of Agreement between Death .Certificate Cancer
Cause and Hc^-oital Diagnosis..
Number %
Complete Agreement 555 75.0
(4 digits in ICOA code) £ 80.7
Agreement to 3 digits 42 5.7
Agreerent to 2 digits 31 4.2
Metastasis entered on D.C. 13 1.8
as underlying
Multiple cancers on D.C. 41 5.5
Primary site not stated
Other 18 2.4
No cancer at autopsy or 20 2.7
biopsy
No records available 20 2.7
Total 740
-------
Table
18
Smoking History in Lung Cancers from Index and
Control Tracts by Sex
Smoking . Non-Smoking Unknown Total
Index Tracts
No.
No.
No.
No
Male
Female
Control Tracts
Male
Female
30
6
63
5
59
67
52
22
2
3
0
0
4
33
0
0
19
0
58
18
37
0
48
78
51
9
121
23
-------
Table 19
Cell Types of Lung Cancer by Census Tract
Oat Squamous Adenocarcinoma Epldemoid Other UK Total
Tract 2303
Male
Ferrale
Other Index
Male
Female
Control
Male
Female
2
-
3
-
12
4
8
1
18
3
57
4
-
1
2
3
16
9
1 1
^ *>
1 1
1
4 3
4
4
-
8
-
18
1
16
2
33
7
no
22
-------
Table 20
Population for Sample and Follow-up
1958-62 1966-67 1968-74
Index Population
Total Deaths 148 72 153
Sample 148 72 153
Hospital request* 96 56 119
to abstract
Control Populaton
Total Deaths 2596 364 1030
Samole 591 258 566
Hospital request* 414 191 441
to abstract
Requests were submitted to all hospitals for those
patients who died in one of the 11 cooperating
hospitals. Some individuals died at home and no
further confirmation could be obtained. Records
were not obtained for all those requested.
-------
isoa.ot
o3 -.ji6i«'
v.. 11« .*'''
\: : /:/
*
70B
mrv
-: (:/:
'2« 04.01.
02
01
105-^:
tSOS.01
INDEX TRACTS
CONTROL TRACTS
MATCH ii :::::
PLANT LOCATION
Figure 1
Map of Baltimore City showing
Index and Control Census Tracts, 1970
PATAPSCO RIVER
....... a)
;2504.0J. f-X*
-------
I960 Baltimore City Line
k
IE2801;-.
!* '..-..
fm '... '-..2718 \...»».
-.:.... 2716 V
2716 '.
\in /.,
'. 1513 \
1308 '
1512 :...-.
..908:.906
S04' /-..-
PATAPSCO RIVER
INDEX TRACTS
I960 CONTROLS
MATCH II CONTROLS ;.
PLANT LOCATION
Figure 2
Map of Baltimore City Showing
Index and Control Census Tracts, 1960
-------
1000-
o
o
o
d
o
LJ
Q_
LJ
a:
o
LJ
o
z
<
I
LU
O
500
§ IOO-
i
UJ
50
20
Comparison of total cancer
rates for census tract 2303
with controls. Age and time
adjusted rates per 100,000
semi-logarithmic scale.
Comparison of Lung cancer rates
Cor Census Tract 2303 with controls
Age and time adjusted rates per 100,00
semi-logarithmic scale.
2303
North
Adjacent
2303
\ V
V /\
North
. Adjacent
Figure 3. ALL CANCER
Figure 4. LUNG CANCER
r 1
5B-9 60-2
I I I l I I
66-7 68-9 70-2 73-4 58-9 60-2
YEARS
i i r i
66-7 68-9 70-2 73-4
0 = less than 5 deaths
-------
SMR
- 2-99
3.00 +
OUTLINE MAP
or
CENSUS TRACTS
1970
BALTIMORE CITY
Fig. 5. Map of Baltimore City showing census tracts with statistically
significant SMRs of 2 or more for white male lung cancer.
(Observed deaths in 1970-1972. Expected deaths based on the
average annual Baltimore City rate. Deaths in 1970-1972.)
-------
wy^f \
W%J^£-3DDtaj^^^J^^X \
fy/y^^L/j^KfJL^to-i w" H"^
Lung cancer death
Other cancer death
<£ Census tract boundar
°x,
\
I: T 1
^. s-
*0
-------
* Lung cancer death
Other cancer death
ft Jy-Jr'.?;i >»i r-r-i>T : r
wyA^LJiJ L3 i-ii^
^^/;^//7>jg^
nl-- L_
'^4! *^=»
sr^SS^xtlS
.'v:
C A
Figure 7
Spot map showing cancer deaths
for 1973-74 by residence at death.
-------
W
NW
6 E
10
sw
II
sw
ANNUAL
SUMMER
LEGEND
E£/ZdVgLOCITIBS 4 THRU 15 MPH
^VELOCITIES 16 MPH AND OVER
VELOCITIES 0 THRU 3 MPH RECORDED AS CALM.
NUMBERS ARE PER CENT OF
TIME WINDS BLEW FROM DIRECTION INDICATED
NOTE: WIND ROSES BASED ON ANALYSIS OF MUNICIPAL
AIRPORT WBO HOURLY WIND RECORDS (1933 THRU 1938)
W 14
SW
WINTER
PER CENT OF TIME
5 0 S 10
WIND DATA
BALTIMORE, MD
Date: March 1950 Drawn by: TIN
Fig. 8. Wind rose for Baltimore.
-------
o r
10 O
C 4J
- -3
O '3 ,
> i
.3 t
JT -i ""^rui r_x
v- £,-'i' 1 i
- '
on
01
i ID
n. i-:
t: G
T5 J
00 00
-------
^"i*'*\ss
^tyv**
r *'-& -., '.. 5,-
f JL_\ i\\
...
t: r r-
cxr-
LT CJ
~--~r O C S.
, V ./i c
H rC c^1^
,^' T- ^J 13
r- - rj
C
> o
O Ti
(J CJ
i- CT> S^
?. ? *
071"
-------
App end ix I
METHOD OF SOIL SAMPLING
The original selection of sampling sites was determined
both by distance from the plant and by direction from north
through south coordinates. We intended to collect about
half the samples within the 1/4 mile radius and 40 per cent
at the next 1/4 mile distance with the remaining samples
collected further out on the radii. Control samples would
be taken from two parks nearby but at a distance greater
than 1 mile from the plant. The field survey team had
problems adhering to the sampling design since the sources
of soil were limited in the area. We attempted to take
samples near residences whenever possible as long as there
were no obvious problems of tree-cover, water run-off, or
redevelopment. For those few samples taken at private
housing, the residents were interviewed concerning the use
of herbicides or pesticides in the area and the sample was
avoided if the soil had been treated. After collection of
the original 101 samples taken at 35 sites under these
directives and including additional samples in the park, a
second set of samples was collected in a north and north-
west direction to determine how far distant the high leve.ls
could be detected. Special emphasis was placed on sampling
from the park which was adjacent to the plant. This park
has a central grassed area which had been recently re-
sodded. Surrounding the park was a dirt-track which had
-------
fence along the plant boundary and near the areas where
railroad cars were filled. Another portion bordered on the
water and the last was adjacent to railroad tracks.
Samples were collected at one, two and four inches at
each location unless otherwise noted. A core sampler, with
a 3.4 inch bore and marked at one inch intervals, was driven
into the ground and samples were removed down to the appropriate
depth marked. Since we had been informed that materials are
often not uniformly distributed in soil, a one foot circle
was marked off around a selected site and a set of samples
i
was collected according to the described technique until the
30 ml. polyethylene sample bottle was filled with soil from
the appropriate depth but from different core samples. This
might require about four separate cores. Any debris such as
rocks, trash, or grass was removed in a funnel prior to
deposition in the bottle. Initially we had tested four
sites using consecutive one inch samples down to a depth of
four inches. We found that samples at three inches were
usually close to those at four and thus it was elected to
take the extreme depth and discard the three-inch level.
Samples in control areas as indicated in table 1 were all
obtained from two city parks, Riverside or Federal Hill.
All sampling was done with polyethylene equipment
except for the corer and the stainless steel spatulas in
-------
order to eliminate possible accidental arsenic contamination.
The metal products used were reportedly free of arsenic and
were the same throughout all sampling. Between each sample,
the equipment was rinsed with tap water and cleaned finally
in 100 mm ethylenediaminetetraacetic acid (EDTA) which was
prepared in glass and stored in polyethylene containers.
The samples were mixed and renumbered before shipment
to the laboratory so that no pattern of numbers by depth or
location could be detected by technicians running the samples,
The analysis was done at Gulf South Research Institute under
the direction of Ms. Mary M. McKown using the digestive
procedure recommended by EPA. All particles 3 mm in size
were removed. The sample was heated with HN03 and HC1 acids
for 3 hours and analyzed by either conventional flame or
flameless atomic absorption spectrophotometry depending on
the level of arsenic.
-------
Appendix II
METHOD OF VALIDATION OF HOSPITAL RECORDS AND PATHOLOGY
OF CANCER CASES
Methods
The deaths for the total 14 years of study were included
in the sample and stratified by control and index census
tract. The sample for hospital record review included all
deaths within the city for residents in the index tracts and
one control tract which was to be used subsequently for the
population survey. All city deaths were stratified by age,
race, and sex and three time periods, 1958-62, 1966-67 and
1968-74. Four control deaths were selected randomly from
each stratum for each index tract death within the same
stratum. For the following analyses no attempt was made to
expand the sample to the original population size. (See
Table 20)
The hospital abstract form was designed to include
information on the following variables:
1. Final diagnosis
2. Source of information for diagnosis
3. Symptoms of cancer
4. Associated medical conditions, especially cancer
5. Arsenic-associated symptoms
-------
6. Personal characteristics as smoking and occupation
7. Family history, especially cancer
8. Laboratory studies including blood, X-rays, and
EKG
9. Description of pathological specimens; operative
or autopsy findings transcribed in detail.
Verification of the identification of the correct individual
on the hospital record was done by name, birthdate, residence,
and date of death.
Records were reviewed in eleven of the hospitals in
Baltimore City. The other five city hospitals either refused
to participate or still had not replied favorably to the
request at the termination of the study. The non-cooperating
hospitals were the smaller ones and did not limit substantially
the number of records reviewed. It became apparent that
problems arise with the scheme of validating patient cancer
records through a review of data from the hospital listed on
the death certificate. On several occasions, the patient
had died at home and no hospital was listed on the certificate.
In this case it was impossible to know where to begin to
seek diagnostic information on the individual. An attempt
to review registry files to identify these missing records
proved to be inefficient. This information could only be
provided by a survey. Another problem was that many individuals
died at hospitals other than the ones in which the cancer
was sparse. An attempt was made to follow these cases
-------
through all hospital admissions in which the cancer may have
been treated but this was not completed in all cases because
of time, money, missing information, or non-inclusion of a
hospital. In these cases we frequently cannot be sure of
the cancer diagnosis. We did not examine the records of
cancers or cancer deaths of city residents at hospitals in
nearby counties. If the study were to be continued, all
outlying hospitals should also be included.
All records were abstracted by two abstractors and were
reviewed by a supervisor for editorial changes. All possible
medical conditions found on record review were listed and
coded by the same nosologist who codes all the death certificates.
The celltypes were classified, in general, according to the
Manual of Tumor Nomenclature and Coding. Since this coding
scheme does not appropriately classify the cells of several
tumors, especially those of non-solid origin, a revision of
the scheme was made to include these cancers if we felt that
their frequency was sufficient to warrant specific classification.
-------
1. REPORT NO.
FJPA 560; i 1 BO-GJ.3
4. TITLE AND SUBTITLE
Lu.ip rarcer/Mortality in Proximity :o a Pesticide Plant
5. REPORT DATE
-March 19SG
6. PERFORMING ORGANIZATION CODE
7. AUTHORS) (jcncv.,eve Matowoski*; Lmamiel Landau0; James Towasda*,
Chrisliana Ui/.ak*; Elizabeth A. Elliott*'; William McEnconr*; Richard Keating
Jysopli Seiftcct-
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
American P'.i'Mic Health Association
1015 ISlU.St., NW e>f
Washirgu-ii, DC 20005 '
10. PROGRAM ELEMENT NO.
11. CONTRACT/GFtANT NO.
68-01-3859
12. SPONSORING AGENCY NAME AND AOOH6SS
. Office ot Toxic1 Substances
U.S. Knvironmentul Protection Agency
401MSI..SW
Washington, DC 20460
13. TYPE Of RE PORT AND PERIOD COVERED
Final Report
14. SPONSORING AGENCY CODE
US. SUPPLEMENTARY NOTES
f
Johns Hopkins L'nivcrsity*
American Public Health Association0
U.S. Environmental Protection Afiincy"1
1G> ABSTRACT
This interim report coven s. study of excess mortality-from cancer in the population reyding neji a chenjiiuil plant in Baltimore
which frail p:ot!uccd iribL'tticidss for 75 years. .
C.'ar.cc: deaths were examined for a fivs .and nine ytai period sufiounding census years to determine net only the absolute mortality
b'Jl '.he changing trends. .
I-'OLI index census tracts incluiiing the one containing the plant were selected based on 1he requirement tliat at lsList.50 pertent of the
ar^a |jy within a three-quartet mile radius of the plant. Ojiiipatison tracts were .-electee! l>jscd on m:,;c'uin° the iiide.x :r?.ct<. by aga,
stTxi snfl socio-economic status. Employee deaths were subtracted. The limy cancrr death rate on an aje adjusted basis hac* bet-n
found lo be_si2nifii;an!ly higher in the eensoi'tract contiiininjt the pl;int. li hod been r]^i^l^. rapidly siiiL-c the mill-sixties A viudy of
thr validation of hospital records and the pathology of cancer cases indicates lha! the excess risk of lun^, cancer cepresmts a r^al ri^k
in mortality and is not the result of local diagnostic and certification practices.
* . '
A corollary study of soil arsenic indicated highest .level? in'the tract with increased lung cancer mortality. The pattern of tilth soil
levels appeared to he related to rail transport of the arsenical material.
further research is required to determine \vhelher other factors mav ulav a role in ,lhc excess of luns .-^ncer.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
Ars'-r,ic
C'Minounity Exposure
Cfni'er
Lur.;; Cancer
Soil Ana!> sis
Uvr.'ith CVrtification
b.IDENTIFIERS/OPEN Ef-JOEO TERMS
i.'pidctnioloitic Study "
Community Study
Canci-1:
LungCancjr
'J'irnstrKnds
I-.nvjronncntal Arsenic
COSATl l-'icld/Group
J1U. DISTRIBUTION STATEMENT
i Unlimited
19. SECUHITV CLASS (This Report)
. NO. OF PAGE:
CLA0
iii page)
. PFIICE
------- |