PB84-199785
Epidemiological Study of the Incidence of
Cancer as Related to Industrial Emissions in
Contra Costa County, California
California Dept. of Health Services, Emeryville
Resource for Cancer Epidemiology Section
Prepared for
Health Effects Research Lab.
Research Triangle Park, NC
Jun 84
U.S. DEPARTMENT OF COMMERCE
National Technical Information Service
NTIS
-------
PESt-J99785
EPA-600/1-84-008
Owe 19S4
EPIDEMIOLOGICAL STTO"? OP THE IIUCIDEHCS 0T CA.TCEE
AS RELATED TO IHTICSIRIAL SttlSSIOTJS 1ST
COSTM COSTA COySTY. CALISTOHtTEA
^7
ttauald 5*. Austin
Tsnw 0, Nslaori
Bis E. Swain
Linda ?. Johnson
California Department of Health Services
Hesourca for Cancer Epidemiology section
^aeiyvllle, CA 945OS
Sract £o. $306396-0\
?toSect Officer
Wilson P.. Riggan
Papulation Studies Division
Health 'Effects Sesearch laboratory
Research Triangle ?ark, FC 277* 1
HEALTH 3FFSCT5 HE52ASCR LASORATOHiT
OFFICE 0? RESEARCH ASS Br?ElCP?®in*
y.s. shvibouhettal protect row assscy
CTNCIimATI OH i'SiSB
-------
TECHNIC*i. SSPCRT 3ATA
rftnf ntti fcsauunr m itc nr-trtt itrc/t fOarrircr.!)
l. "SFOffr --o* eooe
n»c fssf
D.r. Austin, M.D., V. NelSir., &. Swain, L. JcVsQn,
5. im, P. F1es.se!
iSTSosHSriSeSSSTTeST"
LlOfl-
)n. sd^TSACT.-Ofl.-SjMl" ««i.
R806396
OICi^'V^'nQN ^AM« AflO AQ3RBSS
Bept. of Health Servic&s, Resource for Career Epidemiolo^y
Sect'OTi arid Californi-a Turacr Registry, 585C She!Jmcund St.,
5uue 200, tineryvil ie, CA 94508
<3. jffMSfllvS jSENCV hams *nO aco^bss
Health Effects Research laboratory
l\S. Ewirjiprosital Protection Agency
Researcn Triangle Park, N.C. 27^11
' ix ftp* op **ra*rr anq mrioo eavenss
Final: Sept. 1978 ta Dec. 1932
iM,3K«UfiiNiiAa£Na cscs
EPft/eoo/ii
a. sciwLaweNTanY noraa
Pi-eject Officer'. Ulsci E. Rigger, PiO., 3t', -!E3RL
tiaasrsScr
Tiie purpose of this study was to examine the relationship of lung cancer incidence in
Cortra Costa County to asniient lake's cf a.ir pollution. It suspected that the
presence of rss\y irdus-try in the c cmty, ma ily petrochemical plants ird oi" refineries,
could be a contributing factor. Initially, an incidence analysis established t*iat the
Industrial portion of the county had ai excess cr" lung cancer.as compared to the
rei-cifli.ig Kon-industnal portion. Air pollution patterns were stbseqceitly dete^nrnec
by -five permanent air monitoring stations and ten temporary stations w'n'icli monitored
the levels of 12 air pollutants for a period of one year. By correlating the 1970-79
lung cancer rates for each census tract and tract levels of air pollution constituents,
c statistically significant relationship between ambient air SOi and lung cancer in
males, but not in females, was fond, However, when adjusted far the percent cf tfea <-cr- its
poputatian categorized as blue collar, the association was eliminated. Afl i,-iterate* study
of £49 cases and 373 controls was then conducted. Demographic, work history, residential
history, dietary, and smoiniv} htstcry questions comprised the bulk of the data collected.
Analysis indicated that the major contribution to lung cancer in the county was Ave to
ciqarstte smoking, No significant association between lung cancer risk and reasurec
constituents af air pollution was found. Of five broad occupational categories
(indicating possible hazardous exposures} none had any significant relationship tc luf.c
career.
T?u
«V WOffC3 -**D OCCiiweNT ±HA%.fSi3
ft
iistkrjitwHvfvcar^wi jt
saswr* rltidiGicoa i
i?. STirtMsf
please to public
. Ja, SSC'JHITV CLASS ,Oaj,S«WrTj
! UNCLASS1FSED
i ii. no. *aces
I 79
:a. ssojritv clj.ss , ~uj?afwr
WS^USSiFlEQ
zz. '<*ica
-------
NOTICE
This document has been reviewed in accordance with
U.S. Environmental Protection Agency policy and
approved for publication. Mention of trade names
or commercial products does not constitute endorse-
ment or recommendation for use.
ii
-------
FOREWORD
The many benefits of our modern, developing, industrial society are
accompanied by certain hazards. Careful assessment of the risks of existing
and new man-made environmental hazards is necessary for the establishment of
sound regulatory policy. These regulations serve to enhance the quality of
our environment in order to promote the public health and welfare and the
productive capacity of our Nation's population.
The Health Effects Research Laboratory, Research Triangle Park, conducts
a coordinated environmental health research program in toxicology, epidemi-
ology, and clinical studies using human volunteer subjects. These studies
address problems in air pollution, non-ionizing radiation, environmental
carcinogenesis and the toxicology of pesticides as well as other chemical
pollutants. The Laboratory participates in the development and revision of
air quality criteria documents on pollutants for which national ambient air
quality standards exist or are proposed, provides the data for registration
of new pesticides or proposed suspension of those already in use, conducts
research on hazardous and toxic materials, and is primarily responsible for
providing the health basis for non-ionizing radiation standards. Direct
support to the regulatory function of the Agency is provided in the form of
expert testimony and preparation of affidavits as well as expert advice to
the Administrator to assure the adequacy of health care and surveillance of
persons having suffered imminent and substantial endangerment of their
health.
This epidemiologic study assesses the risk of industrial emissions on
the health of the residents of Contra Costa County. Concern about the
health effects of exposure to air pollution generated by the heavily
industrialized area of the county prompted this investigation. Initially
the study focussed on cancer of the trachea, lung or bronchus, comparing the
Industrial and Non-industrial areas of the county and reviewed incidence
cases and rates spanning the years 1975-1979- Subsequently, air monitoring
of the industrial emissions at 15 sampling stations provided data used
herein to calculate census tract specific air pollutant measures for a
correlation analysis of air pollutants and census tract specific lung cancer
incidence. Finally, an assessment of the risk of lung cancer for county
residents was conducted through a case-control questionnaire study linked to
census tract specific air pollution measurements.
F. Gordon Fueter, Ph.D.
Director
Health Effects Research Laboratory
-------
ABSTRACT
This study of the relationship of lung cancer incidence in Contra Costa
County to ambient levels of air pollution was generated by the concern of
public officials and private citizen groups concerned about reports of
elevated lung cancer incidence in the county. It had been suspected by some
that the presence of industrial plants in the county, mainly petrochemical
refineries, could be a contributing factor. The study consisted of five
parts.
First, an incidence analysis established that when the county was
divided into two parts, the Industrial area of the county had an excess
of lung cancer as compared to the remaining Non-industrial area.
More detailed information on the patterns of air pollution in the
county were obtained in the second phase of the study. Five permanent air
monitoring stations and ten temporary stations monitored the levels of 12
air pollutants for a period of one year. These data were incorporated into
later phases of the study.
In the third portion of the study, through a correlation analysis of
1970-79 lung cancer rates and various air pollution constituents, a
relationship between ambient air SO^ and lung cancer in white males, but
not in white females, was found to be statistically significant. However,
the percent of the working population categorized as blue-collar was also
associated with lung cancer in white males and the previous association
between lung cancer in white males and ambient air SO^ levels was
eliminated when this third factor was taken into consideration.
Part four of the study was to have consisted of a linkage of
occupational group cohorts to registry cancer incidence files but was not
conducted for lack of easy availability of occupational group records.
Part five of the study was an analysis of case-control interview data
on a final sample of 622 individuals. Demographic, work history,
residential history, dietary, and smoking history information comprised the
bulk of the data collected.
Analysis of the data indicated that the major contribution to lung
cancer in Contra Costa County was due to cigarette smoking. Further, there
was no identified effect on lung cancer risk contributed by any measured
constituent of air pollution. Of five broad occupational categories
(indicating possible hazardous exposures) none had any significant
relationship to lung cancer.
iv
-------
CONTENTS
Foreword iii
Abstract iy
Figures
Tables vii
Abbreviations and Symbols x
Acknowledgements xi
1 . Overview 1
Introduction 1
Data Source Background 2
2. Incidence Analysis 3
Introduction 3
Methods 3
Data Analysis 5
Conclusions 14
3. Air Pollution Monitoring 17
Introduction 17
Methods 17
Data Analysis 19
Conclusions 26
4. Correlation Analysis 28
Introduction 28
Methods 28
Data Analysis 29
Conclusions 32
5. Occupational Monitoring 33
Introduction 33
Methods 33
Data Analysis 33
Conclusions 33
6. Case Control Study 34
Introduction 34
Methods 36
Data Review 38
Data Analysis 44
Discussion 57
Summary and Conclusions 60
Bibliography 62
v
-------
FIGURES
Number Page
1 . 1970 census tracts, Contra Costa County 4
2. Three year annual age adjusted lung cancer incidence
rates. Industrial and Non-industrial areas.
Contra Costa County, white males, 1970-1978 10
3. Three year annual age adjusted lung cancer incidence
rates. Industrial and Non-industrial areas.
Contra Costa County, white females, 1970-1978 12
4. Contra Costa County incidence analysis.
Cumulative risk of acquiring lung cancer.
White population, 1975-1979 15
5. Locations of air monitoring sampling stations
in Contra Costa County 18
6. Dose-response curves for a composite sample collected
at Brentwood, Ca. , November, 1978-February, 1979 20
7. Contour map of indicated pollutants,
Contra Costa County, November 1978-February, 1979 23
8. Contour map of indicated pollutants,
Contra Costa County, November 1978-February, 1979 24
vi
-------
TABLES
Number Page
1 Five-Year Average Annual Age-Adjusted Incidence Rates for
Cancer of the Bronchus and Lung, Industrial and
Non-industrial Areas of Contra Costa County, 1975-1979 6
2 Number of Incidence Cases of Lung Cancer Cases, Cancer
Deaths and Estimated Population, Contra Costa County,
White and Black Population, 1969-1979 7
3 Number of Lung Cancer Cases by Sex and Year, Industrial and
Non-industrial Areas, Contra Costa County, White Population,
1969-1979 8
4 Number of Lung Cancer Cases by Sex and Year, Industrial and
Non-industrial Areas, Contra Costa County, Black Population,
1969-1979 9
5 Three Year Average Age-Adjusted Rates for Lung Cancer,
Industrial and Non-industrial Areas, Contra Costa County,
White Males and White Females, 1970-1978 11
6 Age Specific Incidence Rates for Lung Cancer, Industrial and
Non-industrial Areas, Contra Costa County, White Males and
Females, 1975-1979 13
7 Cumulative Risk of Acquiring Lung Cancer Expressed as Percent,
Age 30-75, Industrial and Non-industrial Areas, Contra Costa
County, White Males and White Females, 1975-1979 16
8 Analysis of Air Particulate Material by Season, Contra Costa
County, November 1978-October 1979 21
9 Pearson Correlation Coefficients for the Mean Annual Measured
Values for 15 Monitoring Stations and for Computed Values
for 113 Census Tract Centroids, Contra Costa County,
November 1978-October 1979 25
10 Spearman Rank Correlation Coefficients Between Measured
Pollutants for Contra Costa County, November 1978-
February 1979 27
vii
-------
11 Correlation of the Census Tract Specific, Five-Year Average
Annual Age-Adjusted Lung Cancer Incidence Rates in
Contra Costa County for White Males and White Females,
1970-74 and 1975-79 29
12 Correlation of the Ten-Year (1970-79) Average Annual
Age-Adjusted Lung Cancer Incidence Rate to Particulate
Air Pollution Constituents by Individual Census Tract,
Contra Costa County, California 30
13 Correlation Between Sulfates and the 1970-79 Ten-Year
Average Annual Lung Cancer Incidence Rates, by Individual
Census Tract, Contra Costa County, Controlling for Several
Census Tract Characteristics 31
14 Disposition of the 332 Cases Eligible for Interview and
Analysis in the Contra Costa County Lung Cancer
Case-Control Study 39
15 Results of Matching Controls to Cases by Sex and Matching Ratio . . 43
16 Descriptive Statistics for Males. Means and Standard
Deviations for Cases (N=144) and Controls (N=217) 45
17 Descriptive Statistics for Females. Means and Standard
Deviations for Cases (N=105) and Controls (N=156) 47
18 Risk of Lung Cancer, by Sex, Associated with Selected
Environmental Factors in Contra Costa County, With and
Without Controlling for Smoking 50
19 Risk of Lung Cancer for Various Air Pollutants, Controlled
for Smoking, Drinking and Asbestos Exposure, Contra Costa
County Males 50
20 Risk of Lung Cancer for Various Occupational Categories,
Controlled for Smoking, Drinking, Asbestos Exposure and
SO^ Dose, Contra Costa County Males 51
21 Risk of Lung Cancer for Each Factor in the Saturated Model
for Multiple Logistic Analysis, Contra Costa County Males .... 52
22 Risk of Lung Cancer for Each Factor in the Saturated Model
for Multiple Logistic Analysis, Contra Costa County Females ... 53
23 The Risk of Lung Cancer for Males Smoking the Average Dose
and Duration of Smoking Male Cases (N='140) and for Females
Smoking the Average Dose of Smoking Female Cases (N=90) 54
viii
-------
24 The Relative Contribution by Each Risk Factor, Under
Three Assumptions, to the Explainable Proportion of
Male Lung Cancer in Contra Costa County, Expressed as
Cumulative Percents 55
25 The Relative Contribution by Each Risk Factor, Under
Three Assumptions, to the Explainable Proportion of
Female Lung Cancer in Contra Costa County, Expressed
as Cumulative Percents 56
26 Comparison of the Risk of Lung Cancer, by Sex, for Ambient
Air Sulfate Exposure Measured as the Average Annual Level
of the Census Tract of Residence at Diagnosis (SO^ Level)
and as the Computed Total Lifetime Dose in
Contra Costa County (SO^ Dose) 58
ix
-------
LIST OF ABBREVIATIONS AND SYMBOLS
ABBREVIATIONS
AIHL — Air and Industrial Hygiene Laboratory
BAA — benz(a)anthracene
BAAQMD — Bay Area Air Quality Management District
BAP — benzo(a)pyrene
BGP — benzo(ghi)perylene
BSO — benzene soluble organics
CHR — chrysene
COR — coronene
CTR — California Tumor Registry
EPA — United States Environmental Protection Agency
ICD-0 — International Classification of Diseases for Oncology
110 „ — Index of Industries and Occupations
yg/m-5 — micrograms per cubic meter
NCI — National Cancer Institute
ng/m^ — nanograms per cubic meter
NIOSH — National Institute of Occupational Safety and Health
OR — odds ratio
p — significance probability
PAH — polycyclic aromatic hydrocarbons
RCE — Resource for Cancer Epidemiology
RDD — random digit dialing
rev — revertants per cubic meter
S9 — metabolic activator for mutagenic tester strains
SEER — Surveillance, Epidemiology and End Results Program
SF-0 SMSA — San Francisco-Oakland Standard Metropolitan Statistical Area
SIR — standardized incidence ratios
T98 — a mutagenic tester strain
TSP — total suspended particulate
SYMBOLS
NO^ — nitrate
Pb — lead
SO2 — sulphur dioxide gas
SO^ — sulphate
x
-------
ACKNOWLEDGMENTS
This study could not have been carried out without the advice, guidance
and continued support of a great number of people. To properly recognize
each of these people is not possible in such a report. However, certain
individuals must be singled out for special acknowledgements-
The guidance and support of Dr. Wilson R. Riggan, our United States
Environmental Protection Agency (EPA) Project Officer, who provided helpful
advice regarding the administration, funding and scientific review of the
project is gratefully acknowledged. The professional consultation of
several reknowned specialists provided invaluable technical guidance for the
project staff. These persons include: Dr. William Gaffey, Statistical
Consultant, formerly of Stanford Research Institute International;
Mr. James Sandberg, Meteorologist, Bay Area Air Quality Management District
(BAAQMD); Dr. Warren Winkelstein, Jr., former Dean, School of Public Health,
University of California, Berkeley; Dr. Richard Brand, Professor, School of
Public Health, University of California, Berkeley; Dr. Alice Whittemore,
Adjunct Professor, Department of Family, Community and Preventive Medicine,
Stanford University, and Dr. Byron Wm. Brown, Professor and Head, Division
of Biostatistics, Stanford University.
The collaboration of field and laboratory staff of excellent air
pollution analytic laboratories under the direction of Mr. Dario Levaggi,
Executive Director, BAAQMD, and Dr. Jerome Wesolowski, Chief, Air and
Industrial Hygiene Laboratory, (AIHL) Berkeley, and the employment of the
Ames test for mutagenicity by Dr. Peter Flessel, Air and Industrial Hygiene
Laboratory, contributed significantly to the scientific excellence of the
physical and biological measurements used in this study.
The numerous consultations provided by technical staff of the National
Institute of Occupational Safety and Health (NIOSH) are gratefully
acknowledged.
The efforts of a number of members of the Resource for Cancer Epide-
miology Section (HCE) who collected and coded the cancer incidence data
under a National Cancer Institute (NCI) contract are acknowledged, although
their individual mention would be prohibitively long; likewise, the efforts
of a large staff of highly trained interviewers and supervisors, directed by
Dr. Vonnie Gurgin and Ms. Mary Hauck, who carried out the nearly 700 lengthy
and difficult interviews used in this study. The efforts of former RCE
staff, including Dr. William Handel and Mildred Snyder are recognised. The
active collaboration and contributions of Dr. Eva Glazer and Maggie Chiang
are gratefully acknowledged.
xi
-------
The analysis of the enormous amount of data gathered under the auspices
of this project were carried out by a rather large analytic staff, each
contributing in different areas. Key contributions by staff are
acknowledged in the following areas: Susan Lum - computer mapping of air
pollution data; Lynn Palmer - incidence rates; Margaret Hackett - review and
consolidation of cancer case incidence file; Susan Brown - pre-analysis
edits, match/merging of census data in correlation analysis; and Pamela
Chamberlin - census tract coding, tabular presentations, initial incidence
report organization.
A separate note of appreciation is due to Kim Ikeno and members of her
clerical unit for the management of myriad office tasks related to the
conduct of this project, including the preparation of this final report.
Lastly, it is necessary to recognize the efforts of a large number of
people not formally associated with this project. Prominent among them are
former Senator John A. Nejedly and Assemblyman William Campbell, both of
Contra Costa County, whose personal and legislative efforts kept this
project operating. Others who played significant roles include members of
citizens' and environmental groups, the press, concerned members of state
and county government and many citizens of Contra Costa County.
xii
-------
SECTION 1
OVERVIEW
INTRODUCTION
Contra Costa County, located in the northeastern part of the San
Francisco Bay Area, is one of an aggregate of 39 U.S. counties found to have
a high mortality rate for specific cancer sites (Blot et al., 1977). The
fact that the county also has a section which is heavily industrialized,
with five major petroleum refineries and numerous petrochemical plants, and
the fact that 68$ of the total stationary air pollution in the Bay Area
originates from the county (AIHL, 1980a) prompted an epidemiological study
of the incidence of cancer in Contra Costa County funded by the United
States Environmental Protection Agency. The major objective was to
determine whether industrial emissions have a measurable effect on cancer
occurrence. The study was originally proposed by and funds granted to the
Contra Costa County Health Department. Following its award, the grant was
rejected by the Contra Costa County Board of Supervisors and returned to the
EPA with the request that the study be carried out by the California
Department of Health Services. This request was acceptable to both the EPA
and the State of California and the grant was subsequently awarded to the
California Department of Health Services and carried out by the Resource for
Cancer Epidemiology Section of that agency. The study was to have consisted
of five parts:
1. A comparison of the cancer incidence in the heavily
industrialized sections of Contra Costa County to that in the
remainder of the county.
2. Air monitoring, consisting of sampling and chemical analysis of
air to determine the levels of particulate pollution components in
the ambient air.
3. Correlation analysis of lung cancer incidence rates with air
pollution constituents and census tract characteristics.
4. Occupational group monitoring to detect occupational groups at
elevated risk of cancer.
5. A case-control study to identify specific environmental factors
responsible for any excessive amount of cancer incidence in Contra
Costa County.
1
-------
DATA SOURCE BACKGROUND
Since 1972, the Resource for Cancer Epidemiology has maintained a
cancer surveillance system (Linden and Austin, 1974) in the San Francisco-
Oakland Standard Metropolitan Statistical Area (SF-O-SMSA), a five county
area that includes Contra Costa County and which has a base population of
over three million. The RCE is under contract to the National Cancer
Institute to provide cancer incidence data for the NCI Surveillance,
Epidemiology and End Results Program (SEER) (NCI, 1981). An earlier program
of the NCI, the Third National Cancer Survey (1969-1971), also conducted by
the RCE in these same five counties allowed those data also to be used (NCI,
1975) in the analyses which follow.
The RCE cancer data files are maintained by a sophisticated, automated
consolidated cancer information system developed in cooperation with NCI,
and serving as a model for other U.S. cancer registries.
2
-------
SECTION 2
INCIDENCE ANALYSIS
INTRODUCTION
A preliminary analysis of cancer incidence in Contra Costa County for
the period 1972-1975 for all sites combined, lung and bronchus, trachea,
stomach, prostate and lymphoma was completed prior to this analysis for the
Non-industrial area and Industrial area of the county. A line drawn by
subjective means separated the approximate Industrial area of the county
from the rest of the county, constituting the Non-industrial area. The
results of the preliminary analysis showed that a significant difference
existed only for cancer of the lung and bronchus. Therefore, these analyses
were limited to cancer of the lung and bronchus.
METHODS
Lung Cancer Case Selection
Lung cancer cases included for incidence analysis were malignant,
invasive, resident incidence cases for the primary sites of lung, bronchus
and trachea for the period of 1969-1978. Extreme effort was used to
eliminate any erroneous case allocation such as duplicate reports, incorrect
dates of diagnosis, improper census tract assignments, and other factors
such as incorrect demographic and temporal information.
Industrial, Non-industrial Area Definition
In addition, a more objective allocation of Industrial and Non-
industrial census tracts was done. Tracts which were both zoned and used
for heavy industry were assigned to the Industrial area with the remaining
census tracts in the county comprising the Non-industrial area (Figure 1).
Comparison of the lung and bronchus cancer incidence in the two areas for
both sexes and white and black races were conducted, as were temporal trends
for race and sex categories in each area and for the county as a whole.
Development of Population Estimates
Population estimates for 1970-78 by age, race and sex were generated
from census tract data using the 1970 census, a 1975 special census of the
county, and the 1980 census counts of persons by race by using standard
demographic techniques of interpolation and extrapolation. Aspects of the
3
-------
cr*3
fD Q
£TJ
<1
S.n
(D ^
"3
O =>
XJ
*<
1970 census Tracts
&
scale in feet
0 8 000 16000
Figure 1. 1970 census tracts, Contra Costa County*
-------
population denominator file which could possibly contribute to erroneous
results were also ruled out.
DATA. ANALYSIS
Lung cancer incidence rates for the Industrial and Bon-industrial areas
of Contra Co3ta County show a widening difference for each race and sex over
the period of 1969-1978. Final incidence rates for the period 1975-1979
show a significant excess of lung cancer in the Industrial area residents
(Table l). Using the total rate figures of Table 1, the lung cancer rate
for the Industrial area (59.4) exceeds that of the Hon-industrial area
(50.0) by 38-8% or approximately A0%.
Both the number of new lung cancer cases and the number of deaths
increased faster than the population in Contra Co3ta County during the
period 1569-1979 (Table 2). The number of cases aare than doubled during
the period 1969 to 1977- During the last two years the number of cases
appears to have decreased slightly.
Lung can.cer incidence reporting by Industrial/Hon-industrial area of
residence is presented in -able 5 for individual years 1969-1979* While
roughly progressive increases in lung cancer incidence occur for the Non-
industrial area, a peak in cases occurs in the Industrial area in 1975
followed by a gradual reduction in later years.
Though Table 4- demonstrates a numerical difference in the number of
lung cancer cases among black males and females in the Industrial/Non-
industrial areas, a statistically significant difference in rates could not
be demonstrated (see Table l)« The black population is concentrated in two
areas of the county and primarily resides in the Industrial area.
Therefore, the black population would have nearly identical exposures to air
pollution and would make analyses of lung cancer incidence based on
differences in pollutant exposures difficult. For the above reasons, the
study focused on lung cancer incidence in the white population.
?igure 2 presents lung cancer rate data for white males for the
Industrial and Non-industrial areas for the period of 1970-1978. A
difference in incidence rates occurs far the period with a maximum in 1976.
The observed differences were statistically significant at the .01 level in
1972-1976 period and the .05 level in 1977 and 1978 as Table 5 illustrates.
In Figure 3i a trend of high white female rates is indicated in the
Industrial area but far fewer years than in the male comparison. Elevated
rates are present in the 1974-1978 period. However, statistical
significance (at the .01 level) occurs only in the shorter 1975-1977
period, ks in the white male comparison, a decline in rates occurs in the
la3t two years (Table 5)*
In Table 6, white males in the Industrial area had higher age-
specific rates in every age group except the two youngest 0-29 years and
30-34 year3. These differences were statistically significant (at the
.05 level) for ages 50-54 and 55-59. using the difference of means test.
5
-------
TABLE 1 . FIVE-YEAR AVERAGE ANNUAL AGE-ADJUSTED* INCIDENCE RATES FOR
CANCER OF THE BRONCHUS AND LUNG, INDUSTRIAL AND NON-INDUSTRIAL
AREAS OF CONTRA COSTA COUNTY, 1975-79
Group Industrial area Non-industrial area
Males
101 .2
(89.4 -
113.0)**
76.9
(71 .2
- 82.6)**
Females
41 .7
(34.6 -
48.8)
30.3
(27.2
- 33.4)
White males
108.6
(94.5 -
122.7)
77.9
(71 .8
- 84.0)
White females
44.7
(36.5 -
52.9)
31 .0
(27.7
- 34.3)
Black males
92.9
(65.9 -
120.0)
70.7
(40.7
-100.7)
Black females
37.2
(21.5 -
52.9)
15.6
( 3-8
- 27.4)
Total
69.4
(62.7 -
76.1 )
50.0
(47.1
- 52.9)
*Rates are expressed as cases per 100,000 population and are adjusted to
the 1970 U.S. standard.
**Numbers in parentheses are 95$ confidence intervals.
6
-------
TABLE 2. NUMBER OF INCIDENCE CASES OF LUNG CANCER CASES*, CANCER
DEATHS**, AND ESTIMATED POPULATION***, CONTRA COSTA COUNTY,
WHITE AND BLACK POPULATION, 1969-1979
Year
Number of cases
Number of deaths
Estimated population
1969
176
131
555,083
1970
222
165
559,491
1971
236
182
563,909
1972
239
178
568,328
1973
255
177
572,735
1974
271
201
577,148
1975
282
214
581,640
1976
303
203
598,830
1977
359
250
614,171
1978
322
260
629,511
1979
333
233
644,852
*Resident incidence cases with site codes 162.0-162.9 by ICD-0 (World
Health Organization, 1976).
**Deaths among residents with primary cause coded 162 by ICD-0 (World
Health Organization, 1976).
***Population estimated using 1970 and 1980 U.S. census totals and Contra
Costa County census of 1975.
Source: Unpublished SEER data, Resource for Cancer Epidemiology Section
7
-------
TABLE 3. NUMBER OF LUNG CANCER CASES* BY SEX AND YEAR,
INDUSTRIAL AND NON-INDUSTRIAL AREAS, CONTRA COSTA COUNTY,
WHITE POPULATION, 1969-1979
Industrial area Non-industrial area
Year Total White White Total White White
male female male female
1969-1979
663
481
182
1985
1387
598
1969
41
33
8
109
83
26
1970
45
32
13
152
115
37
1971
46
32
14
166
123
43
1972
62
52
10
158
119
39
1973
56
47
9
171
114
57
1974
53
41
12
190
144
46
1975
81
55
26
174
121
53
1976
74
49
25
187
126
61
1977
76
52
24
239
153
86
1978
65
39
26
210
142
68
1979
64
49
15
229
147
82
*Excludes 99 cases for the period 1969-79 where a census tract number
could not be assigned to the address.
Source: Unpublished SEER data, Resource for Cancer Epidemiology Section
8
-------
TABLE 4. NUMBER OF LUNG CANCER CASES* BY SEX AND YEAR,
INDUSTRIAL AND NON-INDUSTRIAL AREAS, CONTRA COSTA COUNTY,
BLACK POPULATION, 1969-1979
Industrial
Non-
•Industrial
Year
Total
Black
Black
Total
Black
Black
male
female
male
female
1969-1979
128
94
34
58
44
14
1969
7
6
1
5
5
0
1970
7
5
2
4
4
0
1971
11
7
4
3
1
2
1972
10
8
2
2
1
1
1973
10
9
1
9
6
3
1974
12
10
2
4
3
1
1975
12
9
3
7
5
2
1976
15
12
3
6
4
2
1977
12
8
4
4
4
0
1978
16
10
6
9
7
2
1979
16
10
6
5
4
1
•Excludes 6 cases for the period 1969-79 where a census tract number could
not be assigned to the address.
Source: Unpublished SEER data, Resource for Cancer Epidemiology Section
9
-------
130
120
110
100
90
80
70
INDUSTRIAL
AREA
NON-INDUSTRIAL
AREA
1
I I I
J I L
1970 71
72 73 74 75 76 77 78
MIDPOINT 3 YEAR MOVING AVERAGE
Figure 2. Three year annual age adjusted lung cancer incidence
rates* Industrial and Non-industrial areas. Contra
Costa County, white males, 1970-1978.
10
-------
TABLE 5- THREE YEAR AVERAGE AGE ADJUSTED RATES FOR LUNG CANCER
INDUSTRIAL AND NON-INDUSTRIAL AREAS*, CONTRA COSTA COUNTY,
WHITE MALES AND WHITE FEMALES, 1970-1978
Average age-adjusted incidence
rates per 100
,000 population
Three
year
White males
White
females
average
Industrial Non-industrial
Industrial
Non-industrial
area area
area
area
1970
74.2 70.4
24.9
18.1
1971
89.7 75.9
26.6
20.2
1972
103.0 74.3
23-9
23-3
1973
112.3 76.5
21 .4
23.1
1974
115.1 75.7
32.5
24.8
1975
114.6 78.3
42.7
24.8
1976
120.3 78.3
50.3
30.4
1977
104.2 79-4
48.1
31 .8
1978
100.0 79-9
40.0
33-9
•Excludes 99 cases where a census tract could not be assigned to the
patient's address.
Source: Unpublished SEER data, Resource for Cancer Epidemiology Section
11
-------
60
50
40
30
INDUSTRIAL
AREA
NON-INDUSTRIAL
AREA
20
10
J L
1970 71
72
73
74
75
76
77
78
MIDPOINT 3 YEAR MOVING AVERAGE
Figure 3. Three year annual age adjusted lung cancer incidence
rates. Industrial and Non-industrial areas. Contra
Costa County, white females, 1970-1978.
12
-------
TABLE 6. AGE SPECIFIC INCIDENCE RATES* FOR LUNG CANCER,
INDUSTRIAL AND NON-INDUSTRIAL AREAS, CONTRA COSTA COUNTY
WHITE MALES AND FEMALES, 1975-1979
Age
White
males
White
females
group
Industrial
Non-industrial
Industrial
Non-industrial
0-29
-
0.2
0.7
0.4
50-34
-
3.6
7.3
2.3
35-44
19-9
11 .9
15.6
8.8
45-49
80.6
52.6
90.6
28.7
50-54
201 .4
61 .8
70.5
55.8
55-59
324.8
147.1
173-0
90.7
60-64
374.0
303-6
192.2
166.5
65-69
421 .3
395-0
167.8
139.9
70-74
605.0
525-7
195.1
143-4
75+
724.4
586.4
137.5
125.6
*Rates expressed as cases per 100,000 population
Source: Unpublished SEER data, Resource for Cancer Epidemiology Section
13
-------
White females in the Industrial area had higher rates in every group
when compared to the Non-industrial area. The difference in rates for
females in the 45-49 and 55-59 age groups was statistically significant
at the .05 level (Table 6).
Figure 4 presents cumulative risk comparison for the 1975-1979 period.
For both white males and white females the major Industrial/Non-industrial
difference in the risk is established by age 60. By age 75, the cumulative
risk of acquiring lung cancer for white males in the Industrial area is
10.6$ versus 7.6$ for males in the Non-industrial area, equating to an
excess risk in the Industrial area of approximately 40$. Comparable risk
figures for white females by age 75 are 4.6$ in the Industrial area versus
3.2$ in the Non-industrial area equating to an excess risk in the Industrial
area of approximately 44$ as shown in Table 7-
CONCLUSIONS
A review of lung cancer incidence data by age, race and sex for various
time periods between 1969 and 1973 shows excess risk in the Industrial area
of Contra Costa County.- Overall the excess is approximately 40$ in the time
period, with a maximum difference occurring around 1976, followed by
narrowing differences in 1977 and 1978.
14
-------
AGE
Figure 4. Contra Costa County incidence analysis. Cumulative risk
of acquiring lung cancer. White population, 1975-1979.
15
-------
TABLE 7. CUMULATIVE RISK OF ACQUIRING LUNG CANCER EXPRESSED AS PERCENT,
AGE 30-75, INDUSTRIAL AND NON-INDUSTRIAL AREAS, CONTRA COSTA COUNTY,
WHITE HALES AND WHITE FEMALES, 1975-1979
White males White females
Age Industrial Non-industrial Industrial Non-industrial
area area area area
35
-
-
.1
-
40
-
.1
.1
-
45
.2
.1
.2
.1
50
.6
.4
.6
.2
55
1 .6
.7
1 .0
.5
60
3.2
1 .4
1 .8
1 .0
65
5.1
3-0
2.8
1 .8
70
7.2
4.9
3.6
2.5
75
10.6
7.6
4.6
3.2
Source: Unpublished SEER data, Resource for Cancer Epidemiology Section
16
-------
SECTION 3
AIR POLLUTION MONITORING
INTRODUCTION
The air pollution study had a number of goals:
1 . To provide data for determining the association, or lack thereof,
of the present cancer incidence rates with air pollution.
2. To determine whether or not mutagenic activity as measured by
the Ames assay could be accounted for by the chemical
characterization of the samples.
3. To develop baseline data on ambient air pollutants for
comparisons with future measurements and for use in future
epidemiological cancer studies.
METHODS
Equipment, Placement and Specimen Collection
A total of 15 hi-volume particulate samplers were strategically sited
at 13 locations in Contra Costa County and two locations in adjacent
counties so as to characterize air quality variations over the entire county
(Figure 5).
Air particulate material was collected on 8" x 10" glass-fiber filters
(EPA Grade Whatman) in standard high-volume samplers which were collected
every sixth day at each of the 15 sampling stations from November, 1978, to
October, 19*79• After sample collection, the filters were weighed to
determine the amounts of total suspended particulate material and delivered
to the Air and Industrial Hygiene Laboratory on a weekly basis. There the
filters were logged in, cut and the pieces distributed for further analysis.
Particulate matter was analyzed for total suspended particulate (TSP),
benzene soluble organics (BSO), sulfate (SO^), nitrate (NO-j), lead (Pb),
selected polycyclic aromatic hydrocarbons (PAH), and mutagenic activity.
Because of the constraints of sample size and resources, the measurements of
PAH and mutagenicity could not be done on each sample. Instead, samples
were composited over the three natural meterological seasons of the San
Francisco Bay Area, viz. Winter: November to February, Spring: March to
17
-------
Figure 5. Locations of air monitoring sampling stations in Contra Costa County
-------
June, and Summer: July to October. For each of the 15 stations, composites
were then prepared by combining all the samples collected over the three
four-month periods.
Chemical and Biological Analysis
TSP, NO^, S04 BSO, Pb
Standard methods were used to analyze for the following five pollu-
tants: TSP was determined gravimetrically (BAAOMD, 1977); NO^ color-
imetrically (AIHL, 1980a); SO, turbidmetrically (AIHL, 1980b); BSO by
Soxhlet extraction (AIHL, 1975); and Pb by wavelength dispersive X-ray
fluorescence (Moore, 1976).
PAH
Specific PAH were separated by high-performance liquid chromatography
(Plessel et al., 1981). Column effluents were quantitated using ultraviolet
absorption and fluorescence. Five PAH were measured: benzo(a)pyrene(BAP),
benzo(ghi)perylene(BGP), benz(a)anthracene(BAA), chrysene(CHP) and
coronene(COR), a possible tracer for vehicle emissions.
Ames Test for Mutagenic Activity
Mutagenicity was measured using the Ames test (Ames, 1975). The Ames
assay was applied according to a two-part protocol. Part one involved
screening the sample in the five standard Ames tester strains both with and
without metabolic activation (S9). The data gave a quantitative estimate of
the mutagenic activity, the most sensitive strain, and the optimum
conditions of metabolic activation for a subsequent quantative analysis.
For the initial screening, five tester strains were used. All samples
showed activity in at least one stain and generally the most mutagenic
activity was measured in one specfic strain (T98). In part two of the
assay, all composite samples were then analyzed using that strain and dose
response curves were obtained. A measure of mutagenicity was obtained from
the slope of the curves. A typical dose-response curve is shown in Figure 6.
DATA ANALYSIS
Overview
The results of analysis of air particulate material for five standard
pollutants (TSP, BSO, Pb, SO^, and NO-z) are summarized in Table 8. The
median and maximum value for each pollutant are listed for winter, spring
and summer. The median and maximum levels of these pollutants were
generally the highest in the winter, reflecting the occurrence of
meterological inversions. The highest level for benzene-soluble organics
(BSO) measured was 41.3 yg/m^ in Concord in December, 1978, and the
lowest was less than 0.8 yg/m^ at several sites in January and February,
1979. The highest Pb concentration was found in Antioch, also in December,
while the highest TSP concentration was in Brentwood in October.
19
-------
VOLUME OF AIR(m3)
Figure 6. Dose-response curves for a composite sample collected at
Brentwood, Ca., November, 1978-February, 1979 •
-------
IJiEVE B. ttSAWSIS OF ilB PASTICOLATE JUtTI-BrAi 31" SS. 79 Jus* 79 Oct 79
Total sBsper.ced
partiCBlE.ta material
t-G
4?
52
229
126
4tS
Bemene soluble DTgaivita
4.8
f .4
S.i
4", .3
32.1
5.1
Lead
0.7
0.2
0.27
0.6
i .a
sa4
6.9
5-3
6.3
19.3
15.9
1 J.2
¦Vs-
7.2
5-*
4.0
3) .6
16,6
\"y.2
*Meiian velc-ss -
-------
Relation Between PAH and Mutagenicity
The correlation coefficient (Spearman) between mutagenicity values for
each of the monitoring stations and the concentrations of the five PAH
measured (BAP, BAA, CHR, BGP, and COR) were 0.33, 0.41, 0.36, 0.31, and
0.51 , respectively, indicating that chemical quantitation of the PAH was not
a good measure of the mutagenicity of the air. This is consistent with the
observation that BAP, BAA, BGP, CHR, and COR represent only about 2% of the
total mutagenic activity of the air. This was demonstrated by comparing the
mutagenicity of the five PAH in amounts proportional to their average levels
measured in the ambient winter air composites with the average mutagenicity
measured in the actual ambient air samples suggesting a poor association
between the ambient air chemical and biological measurements. This
indicated that a more chemical characterization of ambient air is needed in
order to account for the mutagencity as measured by the Ames test.
Geographic Distribution of Air Pollution
In order to correlate cancer incidence data to air pollution
measurements, information about the variations in levels of air pollution by
census tract was needed. Contour maps showing the geographic distribution
of the levels of the seven measured pollutants were constructed using a
computer program called SYMAP (Harvard University Laboratory for Computer
Graphics and Spatial Analysis, 1975). Sampling station coordinates and
associated pollutant levels were used to construct a matrix containing the
pollution levels throughout the county. Contours were then constructed from
the matrix values (Figures 7,8).
Values of the pollutants from each of the 15 sampling stations obtained
were used to compute estimated values for each of the population centroids
for each of the 115 census tracts in the county. The Pearson product moment
correlation coefficients between pollutants were computed for the 113 cenus
tracts (two atypical tracts, a naval base and a retirement community, were
removed). The correlation coefficients between pollutants for the 113
census tracts show very similar relationships to those based on the 15
monitoring stations. While the correlation between Pb and BSO for the 15
monitoring stations was 0.78, the value for the 113 census tracts was 0.70
(Table 9). Likewise, the correlation between BSO and BAA (one of the five
PAH) for the 15 monitoring stations was 0.87, the value for the 113 census
tracts was 0.89- For the fifteen stations, the correlation between BSO
and the two mutagenicity tests (with and without S9) were 0.28 and 0.24
while the correlations based on the 113 census tracts were 0.39 and 0.28,
respectively. Thus, it was felt that the real relationships between various
air pollution constituents, as evidenced from the correlation coefficients
with the data from the 15 stations, were preserved by the computer mapping
technique so that the computed values of the 113 census tracts could be used
with confidence in subsequent analyses using cancer data.
Figures 7-8 show contour maps for Pb, SO^, BSO, the five PAH and
mutagenicity. The geographic distributions for Pb and BSO are similar in
showing that the highest levels are found in a north-south band located in
central Costa Costa County, a region corresponding to the Diablo Valley, a
22
-------
CONTRA COSTA CONTRA COSTA
NOV. 78-FEB.79 NOV 7B~FEB 79
BENZO(qhi)PERYlENE
rxj/tr^
0000 - 999
•illllljiill 1000-1999
;ggS 2 000-2 999
mm 3.000+
BEN^0(q)PYREN£
0000 - 099
!!IM 100- 299
SS 300 - 699
¦¦ 700*
CHRYSENE
r^/m5
0 000 - 099
:ill!!t|;||l 100 - 299
ggg 300- 699
gum 700*
CORONENE
0 000 - 2 999
fflM 3000 - 4 999
mm 5000- 6 999
¦¦ 7 000*
8ENZ(Q)ANTHRACENE
ng/m5
0 000 - 099
!!i::ijiiii 100-299
gggj 300 - 699
¦¦ 700*
MUTAGENICITY ~ S9
rev/m3
0.0- 7 9
illlllllltll a 0-9 9
¦¦ 13 0*
Reproduced from
best available copy.
Figure 7. Contour map of indicated pollutants, Contra Costa County,
November 1978-February, 1979
23
-------
CONTRA COSTA
NOV. 7ft"fEB. 79
iilii 03* 0 8
liH 0 8' 12
OflGANlCS J
0 0< 2 0
«a 2 0<40
mii to«8o
SHI 8 0*15 0
sulfates Udjl"'
00* 4 0
1«® 4 0*50
IBIS 5 0*70
ma 7 0* 9 0
Figure 8.
Contra Costa County,
fflao of indicated pollutants, Cont
Contour map or 7g
November 1978-February,
Reproduced from
best a valla hip mrr
-------
TABLE 3, PEARSON COBBE1ATIOE' COEFFICIENTS FOR THE MEM AiBTJAL MEASURED
VALUES FOR 15 MONITOR DIG STATIONS AITC) FOR CO'TPUTED VALUES FOE 113 CEHSOS
THACf CEKTP.013S, C0>)T5A COSTA CCCTTY, gOVBiOEH 1973 - QCT03BS 1979
Comparison 15 Stations 113 Census tracts
BSO
vs
Pb
.78
.70
BSO
V3
Hut (-S9)
.24
.28
BSO
VS
Hut (+S9)
.28
• 39
BSO
vs
EAA
.87
.89
BSO
vs
EAP
.SO
.84
BSO
vs
BGP
.64
.71
BSO
vs
CHR
¦ S3
.84
BSO
vs
so4
.19
.17
TSP
va
NO-j
.64
• 75
25
-------
natural pollution sink and the path of a freeway. The Pb map is consistent
with the fact that the largest source of Pb in the area is the automobile.
Comparison of the BSO and Pb maps suggests the contribution of the
automobile to the BSO levels may be significant. The SO^ distribution
differs from the Pb distribution by running in an east-west direction along
the industrial belt. This is consistent with the fact that sulphur dioxide
gas (S0p)» the precursor of S04, is emitted by stationary sources,
primarily chemical industries, refineries and power plants, all located
along the industrial belt. The patterns of the five PAH are similar to one
another and to lead.
Sources of PAH and Mutagenicity
Table 10 gives the correlation coefficients among the pollutants
measured. Values are high between Pb, BSO, and the five PAH, implicating
vehicles as the source of these pollutants. However, the correlation
coefficients between mutagenicity and Pb, BSO, and the five PAH are small
suggesting that mutagenicity concentrations may come from multiple sources.
Information can be obtained on sources by the use of simple ratio
techniques. The ratio, for example, between BAP and BGP for a number of
combustion sources have been established. Automobiles tend to have the
lowest ratio, O.J to 0.4 while other sources tend to be equal or greater
than 1. The ratio found in winter in Contra Costa County ranged from 0.08
to 0.25, consistent with the automobile being a major contributor of BAP.
The high correlation between Pb and PAH, the similarity of Pb and PAH
contour maps, and the value of the ratio all point to the automobile as a
major contributor to PAH.
CONCLUSIONS
The Ames test can be carried out with enough precision to make it a
useful tool for determining trends, geographic distributions, or temporal
variations of the mutagenic potential of ambient air particulate matter.
It is advantageous to use both the Ames test and chemical character-
ization together in attempting to predict the potential carcinogenic effects
of ambient air particulate matter.
In order to account for the mutagenicity as measured by the Ames test,
a more complete chemical characterization of ambient air particulate matter
will be needed.
In Contra Costa County, mobile sources are significant contributors to
PAH; however, more research is needed to define the major sources of
particulate mutagens.
26
-------
TABLE 10. SPEARMAN RANK CORRELATION COEFFICIENTS BETWEEN MEASURED POLLUTANTS
FOR CONTRA COSTA COUNTY, NOVEMBER, 1978 - FEBRUARY, 1979
Pb BSO BAP BAA CHR BGP COR +S9 -S9 TSP S04 N05
Pb 1.00
BSO
.87
1 .00
BAP
.84
.92
1 .00
BAA
.84
.94
.90
1 .00
CHR
.82
.96
.92
• 94
1.00
BGP
.88
.87
.93
.92
.88
1 .00
COR
.87
.82
.88
.82
.80
.91
1 .00
+S9
.44
.51
.33
.41
• 36
• 31
.51
1.00
-S9
.46
.59
.39
.44
.47
.34
.42
.88
1 .00
TSP
.33
.58
.39
.65
.59
.48
• 35
.45
.48
1 .00
so4
.39
.26
.31
.20
.31
.37
.47
.27
.25
.07
NO,
.14
.20
-.04
.23
.21
.12
.06
.39
.42
.71
-------
SECTION 4
CORRELATION ANALYSIS
INTRODUCTION
A correlation analysis between cancer incidence rates with air
pollution constituents and census tract characteristics was conducted to
establish any relationships which might suggest a cause of the excess of
lung cancer found in the Industrial area residents and to formulate
hypotheses which could be tested through the case-control study.
METHODS
Case Selection, Population Estimates and Socio-economic Variables
The following information was available for the correlation analysis:
1. Estimates of air pollution data.
2. Refined population estimates by age, race and sex.
3. Number of lung/bronchus cancer cases by age, race and sex.
4. Demographic information (i.e. socio-economic variables) obtained
from the 1970 U.S. census and the 1975 Contra Costa County census.
5. Ten and five-year average annual age-adjusted lung cancer incidence
rates, by sex and race based on (2) and (3) above.
Correlation Methods
Pearson product moment correlation coefficients were computed for
comparisons between estimated annual values for each air pollutant
constituent for each census tract and the 5- and 10-year average annual age-
adjusted incidence rates for cancer of the lung for males and females
for each census tract. Certain census tract-specific socio-economic
variables from the 1970 United States census and 1975 Contra Costa County
census were also correlated with cancer incidence and air pollutants.
Partial correlations were computed to control for these attributes.
28
-------
TABLE 11. CORRELATION OF THE CENSUS TRACT SPECIFIC, FIVE-YEAR AVERAGE
ANNUAL AGE-ADJUSTED LUNG CANCER INCIDENCE RATES IN CONTRA COSTA COUNTY FOR
WHITE MALES AND l/HITE FEMALES*, 1 9W-74 AND 1 975-?9
White males
White males White females
White females
1970-74
1975-79 1970-74
1975-79
White males
1970-74
1 .00
.
White males
1975-79
.17
1 .00
—
White females
1970-74
- .02
o
o
-
White females
1975-79
CM
.54 (p<0.0l) .05
1 .00
""Correlation of the 10-year rates for white males vs
white females = .50 (p<0.000l) (N = 113 census tracts)
DATA ANALYSIS
Since the correlation of census tract-specific average annual lung
cancer rates for the 5-year periods, 1970-1974 and 1975-1979, for white
males and white females differed significantly from zero for only two out of
the six possible correlations, only ten-year rates were used (Table 11). No
statistically significant differences were demonstrated for the black
population owing possibly to the sparse geographical distribution of the
black population mentioned in the incidence analysis. The correlation
between the ten-year census tract-specific lung cancer incidence rates for
white males and females was O.JJ (p<.000?) for all tracts and 0.50 (p<.OOOl)
with the two atypical tracts removed (a naval base and a retirement
community).
The ten-year census tract-specific lung cancer incidence rates were
significantly correlated with only one measure of particulate air pollution,
that of S0< (Table 12). The correlation coefficient for white males was
0.46 (p<.000l) and for white females the correlation coefficient was 0.16,
not significant. Controlling for the percent of households in the census
tract in which the head of the household had resided in the unit for 20 plus
years, as determined in the 1975 Contra Costa Census, did not effect the
correlation. Controlling for the percent of census tract residents of
Spanish origin reduced the correlation slightly for males (Table 13).
Controlling for the percent of blue collar workers significantly reduced the
correlation between SO^ and lung cancer for males. This reduction was
29
-------
TABLE 12. CORRELATION OF THE TEN-YEAR (1970-79) AVERAGE ANNUAL AGE-ADJUSTED
LUNG CANCER INCIDENCE RATE TO PARTICULATE AIR POLLUTION CONSTITUENTS* BY
INDIVIDUAL CENSUS TRACT**, CONTRA COSTA COUNTY, CALIFORNIA
Air pollution vs Average annual incidence rate (per 10^)
constituent White males White females
Mutagenicity
(-S9)
-.01
-.07
Mutagenicity
( + S9)
.05
-.01
BSO
.11
-.09
BAA
-.05
-.18
BAP
-.09
-.18
BGP
-.13
-.16
CHR
-.03
-.14
TSP
.20
(PC0.05)
-.07
Pb
.11
.06
NO^
-.09
-.24
so4
.46
(P<0.0001)
.16
*Mean annual value, November 1978 - October 1979
**113 tracts
30
-------
TABLE 13. CORRELATION BETWEEN SULFATES AND THE 1970-79 TEN-YEAR AVERAGE
ANNUAL LUNG CANCER INCIDENCE RATES, BY INDIVIDUAL CENSUS TRACT*, CONTRA
COSTA COUNTY, CONTROLLING FOR SEVERAL CENSUS TRACT CHARACTERISTICS**
Controlling for: SO^ vs SO^ vs
White males White females
1.
Nothing
.46
O
O
O
.16
2.
Head of household resident of
census tract for 20+ Years
.46
(p<.0001)
• 13
3.
Percent Spanish origin
.42
(p<0.001)
.19 (p<0.05)
4.
Percent blue-collar workers***
.21
(p<0.05)
.02
a) Percent unskilled laborer***
.29
0
•
0
.04
b) Percent skilled Laborer***
• 29
(p<0.01)
.09
5.
Percent of households below poverty level
• 30
0
«
0
0
.07
6.
Median family income
.21
(p<0.05)
-.03
7.
Median school completed by head of
household
.17
-.06
*113 census tracts
**Data from 1975 Contra Costa County special census
***Data from 1970 U.S. census
31
-------
from independent effects of both the skilled and unskilled laborer
components of blue collar workers. Controlling for the percent of
households below the poverty level reduced the SO^ lung cancer correlation
only moderately. Controlling for education or income variables reduced or
destroyed the correlations for males (Table 13).
CONCLUSIONS
Several conclusions can be drawn from the correlation analysis in
association with the data from the incidence analysis and air monitoring
portion of the study.
It is apparent that an excess of lung cancer exists in the Industrial
area of the county. This excess has developed over the past decade.
Mutagenic activity, as measured by the Ames test, is identifiable in
the particulate matter in ambient air in Contra Costa County and is best
associated with the distribution of BSO and Pb, suggesting mostly automobile
sources.
The ten-year occurrence of lung cancer in white males is only weakly
but significantly associated with the distribution of SO^ particulate
matter in the ambient air. This association is reduced equally by
controlling for the percent of skilled, or unskilled, laborers in each
census tract. Controlling for both factors combined nearly destroys the
observed association.
The source of SO- in the air is known to be almost exclusively from
oil refining, chemical manufacture or oil combustion for electrical power
generation. The association of lung cancer in males with this factor might
suggest a carcinogenic effect from past petrochemical emissions were it not
for three factors:
1. The current mutagenic activity in the particulates is distributed
differently.
2. The association does not occur in females.
3. The association appears to be mediated through a residential
pattern of blue collar workers which correlates well with the
distribution of SO^ (0.67; p<0001). Although these data
permit no conclusions regarding the causes of lung cancer in the
county, they suggest that a significant contribution to
incidence by blue collar workers residing in areas of close
proximity to petrochemical plants.
32
-------
SECTION 5
OCCUPATIONAL MONITORING
INTRODUCTION
The objective of this project was to determine the identification
of occupational cohorts at increased risks of cancer.
METHODS
A data base was available from an occupational surveillance system,
based on files from the RCE first prepared in 1976, consisting of membership
rosters of various unions in the San Francisco-Oakland SHSA. A cohort,
composed of approximately 6,400 union members who were residents of the San
Francisco-Oakland SMSA, was selected from these data- Six occupational
groups were represented; asbestos workers, bakers, painters, plasterers,
plumbers and roofers.
The basic methodological approach used in this study consisted of
computer matching the cohort of occupational workers to the master file of
cancer cases maintained by the California Tumor Registry (CTR). The
observed cases for each cohort and primary site were then compared to
expected cancer cases which were calculated using age-sex-year specific
incidence rates for the SMSA. Standardized Incidence Ratios (SIR's) were
also estimated.
DATA ANALYSIS
Increased cancer incidence was demonstrated among asbestos workers, an
occupation widely recognized as having high risk for respiratory and
gastrointestinal cancers.
CONCLUSIONS
This part of the study was not continued due to the lack of cooperation
of additional sources of cohorts, i.e., employees and unions. Since this
effort was consuming an inordinate amount of project resources it was
discontinued.
33
-------
SECTION 6
CASE CONTROL STUDY
INTRODUCTION
The Relationship of Lung Cancer Incidence in Contra Costa County to Air
Pollution
The over-all goal of this part of the study was to identify specific
environmental factors responsible for lung cancer in Contra Costa County.
The particular objectives were twofold: first, to identify any-
significant environmental factors which contribute to lung cancer
incidence in that county and second, to specifically evaluate whether air
pollutants, released into the ambient air, have had any effect on the
observed incidence of lung cancer in the community.
Preceding components to this project generated findings to be addressed
by this component. The incidence analysis established that when the county
was divided into two parts, the Industrial area of the county had an excess
of lung cancer as compared to the remaining Non-industrial area. Further-
more, this excess had increased significantly during the decade of the
1970's to a peak in 1976, after which a decline in the excess was apparent.
Through a census tract specific analysis it was further demonstrated that,
although an incidence rate excess of approximately 40$ was found in the
Industrial area of the county, considerable variability existed in both
areas, so that significantly high and significantly low rates of lung
cancer were found in census tracts in both areas. Thus the Industrial area
was a mixture of census tracts of high, medium, and low rates, the average
of which exceeded the overall rate of census tracts comprising the Non-
industrial area.
Through a correlation analysis of 1970-79 lung cancer rates by census
tract and various air pollution constituents only one statistically
significant relationship was found to exist. That relationship was between
ambient air SO^ and lung cancer in males, but not in females. However,
the percent of the population categorized as blue collar workers was also
associated with lung cancer in males and the previous association between
lung cancer in males and ambient air SO^ levels was considerably reduced
when this third factor was taken into consideration.
34
-------
The results of the indirect studies such as the correlation analysis
summarised above, led to the ease-control study of individuals described
below.
Literature Review
Respiratory cancer rates vary with geographic location but are still
increasing over time (Doll,1981). For the period "9"3-!977 the age-adjusted
incidence rate in U.S. males was 78.3/100,000 for cancer of the lung and
bronchus. In females, the corresponding rate was 22.2/100,000.
In the San Francisco Bay area, the age-adjusted incidence rates for
lung cancer during this period were higher; 83-2/100,000 in males and
£9-8/100,000 in females (Young,1981).
Overall survival rates for lung cancer are so poor that incidence and
mortality rates are often used interchangeably. In the United States and
most other developed countries cancer of the lung presently is responsible
for almost 40 percent of male cancer mortality (del Regato,1977). Yet in
1923, the Kassachusetts age-adjusted mortality rate/100,000 was 1.55 for
males and 1.50 for females (Lombard,1958). The increase, which has taken
place nairJLy during the last forty years, has been related to cigarette
smoking {flamaond, 1975)-
U.S. men began ar.cking cigarettes in appreciable numbers at the turn of
the century. The habit became more widespread after fcTorlc War I. ar.d
reached a peak of 55 percent in 1937 {Beamis,1975;Eurbank,1972)* It aegas
to decline in 1957 to about 42 percent in 1970 and 39 percent in 1975
(National Clearinghouse,1975)¦ For women, the rise began during World War
II, reaching a high of 39 percent in 1961 , then falling to 30 percent by
1970 and 29 percent in 1975 (Beamis,1975)¦ In male heavy smokers the risk
for lung cancer has been established at about ten times that for non-smokers
(Hammond,1975)• Third National Cancer Survey interview data indicate that
for women who smoke heavily, the risk may be as high as sixteen times that
for non-smokers. Cigar and pipe smokers have an approximately twofold risk
of developing lung cancer (Williams,1977).
Burbank (1972), using an estimated latent period of thirty years
between first exposure to smoking and tumor development, has fitted both
male and female lung cancer mortality rates from 1950-1968 onto a tobacco
dose response curve based on estimates of past cigarette consumption.
There has also been considerable speculation on the extent to which
general air pollution contributes to lung cancer mortality in urban areas.
Clemmesen (1977J has suggested that it may have a non-specific effect on the
bronchial mucosa to aake it nore scsceptibla to carcinogens such as tobacco
saoke. ?!any pclyaromatic hydrocarbons IPAH; are known to be carcinogenic
and the most commonly used indicator for PAH is benzo(a)pyrene concentration
(Pike, 1975) a known carcinogen directly measured in the present project.
More recently, Vena (1982) reported a study of the effects of air
pollution in Erie County, Pennsylvania, using total suspended particulates
35
-------
as a measure. He found evidence suggesting a relationship between air
pollution and smoking and possibly between air pollution, occupation and
smoking, for 50 or more years of exposure to air pollution.
Evidence on other pollution components is scanty. Higgins (1977) found
a correlation between sulfates and lung cancer in fifty SMSA's. Blot and
Fraumeni (1975), also in a correlative study, found excess lung cancer in
counties that had arsenical air pollution from non-ferrous metal smelters
and refineries.
METHODS
Case-Control Study Design
Case-control studies are a standard tool of epidemiological research.
Simply put, a group of persons having a disease (cases in this study, lung
cancer) is compared with a group of persons not having the disease
(controls). In matched case-control studies, the controls are carefully
selected so that they are similar to cases in important background
variables. In this study, cases were matched to controls of the same race
and sex, and similar age. The study populations of cases and controls are
interviewed to obtain detailed information on study variables pertaining to
each participant. The data are then subjected to statistical analyses to
detect significant differences in known or suspected risk factors for
contracting the disease. The outcome of these analyses is the product of
the study. The report typically identifies risk factors which predict the
probability of any individual in the study being a case. Vhen these risk
factors fulfill the criteria for causal factors their relative importance as
a cause of the disease under study can be directly determined.
Questionnaire Design
The questionnaire was developed to conform to standard survey research
practice. The questions were designed to be read verbatim and were
extensively pretested to eliminate any ambiguities. Question order, probes,
and skip patterns were organized so that the same instrument would be
administered to each respondent. Care was taken to format the questionnaire
for ease of administration, in the interest of both accuracy and of the time
required to record the answers. The questionnaire was pilot-tested on a
small sample of respiratory cancer patients and healthy residents of Alameda
and Contra Costa counties who were not included in the study.
A manual of detailed interviewer instructions was prepared. This
included question-by-question specifications for the questionnaire.
To ensure the accuracy of the data collection and to minimize refusals
by cases and controls, we employed interviewers, trained in survey research
interviewing techniques, for administration of the questionnaire ana
selection of controls. An interviewer training session for the study was
held before the start of the field work and at intervals during the study
period. This included mock interviewing during the briefing and practice
interviews in the field. These were reviewed by the field work supervisor
36
-------
and, where indicated, fjethsr trailing wag conducted.
Questionnaires were edited and reviewed by the supervisor before- being
coded and sent far Key data entry.
Case/?rogy Selection
kll cases of cancer of the trachea, bronchus or lur.g among black or
white residents of Contra Costa County, newly diagnosed between Hay 3, 1980
and July 31, 19S1, and who were at least 35 years of age and leas thac 75
years of age at diagnosis comprised a group of 532 eligible cases.
Some eligible lung cancer patients were too ill to participate in the
3tady or, in other instances, had succumbed to their disease before an
interview could be conducted, Whenever passible, a proxy for the case was
identified for interviewing. Moat proxies were the spouse or other
immediate family member of the case but in a few instances the proxy was a
close friend of the case,
floatral Selection
To meet a matching requirement of at least one control per case, the
required number of controls in each of 32 age, race, and sex strata was
computed based on the percent distribution of the cases- The control
population was therefore a representative sample of that non-cancer
population of Contra Costa County with an age, race, and sex distribution
equivalent ta the lung cancer cases. To assure an ad-aquate number of
controls, each stratum was slightly overfilled to allow for possible
subsequent deletions necessitated by insufficient respondent-supplied
detail-.
A target number of 350 -controls for the cases was selected- Because so
many variables were under consideration, each with a different prevalence in
the study population, it is not possible to present a single "beta error" or
a minimum required sample size- ?or most variables, a sample of 250 cases
and 350 controls is adequate to detect relative risks of zvc or greater, at
a significance level of = 0.05 and a probability of detection of &3% or
greater (Oliphant, 1931).
Random Digit Dialing
Controls were selected from the general population of Contra Costa
County by a random digit dialing C-EOD) technique. Briefly, this technique
required obtaining froE the telephone company telephone prefixes in use
in ~he area tcj be sar/pl-ec. "To selecT a nunoer to be dialed, a telephone
prefix was first selected by a random process front among the prefixes
serving the sampling area. Then by a second random process the auffix was
chosen fran among all possible suffixes to the selected prefix, -his
number, which may Iia/s been a business number, an jaassigned number or even
37
-------
an unlisted number as well as a listed residential number was called to
select an individual for the study. Business numbers were not eligible for
selecting controls. For non-answered calls to working numbers, repeat calls
were made on a specified schedule which assured at least six calls; two on
weekday mornings, two on weekday evenings and two on weekends, until either
an answer was received or six calls were placed. For calls made to a
residence, & census of all household members was taken and from among the
eligible members (if any) one person was selected, also by a random process,
and invited to participate in the study.
The possibility of bias by the HDD technique exists if different
proportions of controls had multiple phones or had no phones through which
they could be reached. Because all cases had telephones it was permisaable
to limit control selection to those who could also be reached by phone. A
small proportion, approximately the same among cases and controls, had two
or more telephones in their household.
Other sources of bias, not peculiar to this control selection technique
but which are hazards to the general process of control selection in case-
control studies, include a "response" bias. This bias can occur in several
ways. It is possible that proportionally more (or fewer) cases than
controls agree to participate in the study and that those who refuse to
participate differ from the participants in some way that is related to one
or more variables under study. In this study, another source of response
bias was possible. Thi3 was the possibility of different response rates in
different areas. In fact this was the case and will be discussed in a later
section.
DATA REVIEW
Final Case-Control Counts
After key data entry, the process of editing the file began. Each
variable was checked for completeness and interfield comparisons for
consistancy and accuracy were performed. At the beginning of this process
there were a total of 268 cases and proxies and a total of 410 controls.
Twenty-six respondents were found to have incomplete data fields and were
deleted. Fourteen of the 19 cases deleted due to incomplete data were
proxies. In addition, 7 controls were deleted due to incomplete data.
There were a few age, race and sex combinations in which no cases
occurred during the 3tudy period, even though one or two were anticipated.
Controls collected for those particular combinations were dropped.
Altogether, thirty controls were dropped for a lack of corresponding cases.
At the end of the editing and matching processes 19 cases and 37 controls
had been deleted leaving 249 cases and 373 controls for analysis. Through
interviewing either proband or proxy cases, 75% of the eligibles were
available for analysis. These results are summarized in Table 14.
38
-------
TABLE 14. DISPOSITION OF THE 332 CASES ELIGIBLE FOR INTERVIEW
AND ANALYSIS IN THE CONTRA COSTA COUNTY
LUNG CANCER CASE-CONTROL STUDY
Probands Proxies
Eligibles 238 94
Refused participation 12
Dead, no proxy available 11
111 5
Moved 11
Other 25
Incomplete questionnaire 5 14
Total deletions 69 14
Interviewed and analyzed 169 80
Generated Variables
Additional variables were generated from information collected in
the questionnaire or in combination with other data concerning air
pollution, residential water supply, or county population estimates.
Air Pollution
The measure of the respondent's exposure to air pollution was
expressed as an estimated cumulative dose in Contra Costa County for each
pollutant. In the air monitoring component of this study, census tracts
in Contra Costa County were assigned air pollutant values based upon
measurements taken at various monitoring stations throughout the county
and constructed by interpolation of air pollution station measurements.
Air pollutant values for each census tract were those values at each
census tract population centroid. A total air pollutant dose for each
study participant was calculated from the length of residence in each
census tract. Thus, an air pollutant dose is the sum of the products of
the duration of residence in each census tract times the corresponding
air pollutant values for each census tract. Because the purpose of this
analysis wa3 to evaluate the effect of air pollution in Contra Costa
County on lung cancer incidence, residence time outside the county was
not given any value. This approach does not create any bias although
recent immigrants to the county contribute little information to the
analysis.
39
-------
Smoking
The respondents' smoking experience was characterized by several
parameters; total smoking duration, total pack years and average packs
smoked per day. The respondents' smoking history was recorded by separate
smoking periods. Each period was defined by the number of cigarettes (later
converted to packs) smoked per day and the number of years of smoking at
that level. Thus, total smoking duration is the sum of the durations for
each smoking period. Pack years is the sum of the products of the number of
packs smoked per day and the duration for each period. Average packs smoked
per day is pack years divided by total smoking duration.
Occupation
The analysis for occupational exposure associations was based on the
classification of occupation and industry titles in the 1980 Alphabetical
Index of Industries and Occupations (110), (Bureau of Census,1980).For
each work experience, an occupation and industry code was assigned.
The employment groups considered to be at risk of possible exposure to
hazardous substances in the analysis were derived from the occupations
listed under the 110 classifications of precision production, craft and
repair occupations, operators, fabricators, and laborers. Four major
industrial groups: petrochemical, metal, construction and all "other"
remaining industries were also created.
"Construction" is a single industrial classification. "Petrochemical"
industries include the manufacture of chemicals and allied products,
petroleum and coal products, and rubber and miscellaneous plastics
products. "Metal" industries include metal raining, primary metal industries
such as blast furnaces and foundries, metal product manufacturing, and
machinery manufacturing including electrical and transportation equipment.
The remaining group of industries, "Other", includes agriculture; the
manufacturing categories of food and tobacco, textiles, paper and printing,
leather, lumber and furniture, stone products, and professional equipment;
the transportation, communication and public utilities industries; wholesale
and retail trade; repair and personal services; and professional and public
administration.
Over the entire employment history, if the worker fell into an
indicated employment group and industry group, the duration of time worked
in the industrial group was calculated and accumulated.
An asbestos exposure variable was created from various occupational
categories. All shipyard occupations plus all other jobs for which asbestos
exposures were reported by the respondent were combined to form a total
duration of asbestos exposure per respondent. This variable is not mutually
exclusive of other occupational categories. No attempt was made to quantify
individual asbestos exposures other than as job duration.
40
-------
Water Source
The distribution of the water sources in Contra Costa County were
available from the Sanitary Engineering Branch of the Health Services
Department. The three main water sources for Contra Costa County are
Mokelumne River, San Pablo Reservoir, and Sacramento River Delta water
systems. The census tract population centroid and the distribution of the
water sources were used to assign a water source to every census tract. In
turn, each participant was assigned a water source based on the census tract
of residency at the time of interview or, for cases, diagnosis.
Toxic Waste Dumps
Certain of the census tracts in Contra Costa County contain known dumps
of toxic or chemical waste. These census tracts were combined to form a
toxic waste dump area for comparison to all other tracts.
Sample Adjustment Factor
In addition to matching the controls to the percent distribution of
cases, they were also selected to represent the geographic distribution of
the non-cancer population. To evaluate possible response variation, the
number of controls expected from each census tract was computed and compared
to the number actually obtained. It was found that one area of the county
was overrepresented among controls and a separate small area of the county
was underrepresented. Therefore the responses of the controls from those
two areas were appropriately weighted in analyses to eliminate any effect
of the response bias.
To assess the results of random sampling for controls within age, race,
sex categories, the observed results were compared with expected numbers
computed as follows.
For each age group (under 50, 50 and over), race (white, black), sex
(male, female) category (k), the expected number of controls for a census
tract was computed as the proportion of the population of category k
residing in census tract "x" times the total number of sampled controls
observed in the study for category k. Mathematically this can be expressed
p
t? = xk * n
xk rro k
k
where, = expected number of controls for tract x, category k.
^xk = estima'te
-------
Summing across all (k) categories for a given tract yielded the total number
of controls Ex expected for tract (x).
8
Ex = 51 Sxk
k=1
where, Ex = total number of expected controls for tract (x).
An over- or under-response rate of observed controls in the tract was
said to exist if the statistical quantity, chi-square,
2
? (o-e)
X = exceeded 3-5 •
The result of this comparison showed an over-response rate in 13 census
tracts of the Richmond-San Pablo area of the county. The ratio of observed
to expected in these tracts was as much as 9 times greater than expected.
An under-response rate, as low as one-tenth of expected was found to have
occurred in two census tracts in Concord, one in Antioch, one in the
Brentwood area, and one in Pinole (see Figure 6).
An adjustment factor variable, the observed divided by expected, was
therefore assigned for each control meeting the above criteria. This factor
was incorporated in all statistical analyses to adjust for the variable
response rate.
Alcohol Intake
Alcohol consumption was estimated on the basis that the amount of
alcohol in an average glass of wine, a can of beer and a jigger of spirits
or liquor is approximately the same. The number of each of these drinks
consumed per week was added to form a total number of servings per week,
treated as a continuous variable.
Matching Within Age, Race, and Sex Strata
For the purpose of subsequent statistical analysis, i.e., multiple
logistics regression with matched case-control data, it was necessary to
stratify the case-control population such that each case was matched to
controls with identical race, identical sex, and approximately equal age. A
variable matching ratio was used.
42
-------
TABLE 15- RESULTS OF MATCHING CONTROLS TO CASES
BY SEX AND MATCHING RATIO
The number of (249) cases matched to the indicated
number of (573) controls
Matching ratio
Sex 12 3 4 5 Total
Male
80
58
4
1
1
144
Female
65
29
11
0
0
105
Total
145
87
15
1
1
249
Within a particular age, race, sex stratum an algorithm for matching
cases to controls by age was developed. Within a stratum cases and controls
were separately ranked from low to high by ascending age. The number of
controls assigned to a case was given by the following formula:
No. of controls for case (i) = INTEGER(.5+ CR^ * MR) -
INTEGER( . 5+ CRi_1 * MR)
where, CR^ = case rank number,
MR = matching ratio = (No. of controls in cell)/
(No. of cases in cell)
INTEGER = use the number to the left of the decimal
result of the computation within parentheses.
For the first case, CR^_^ = 0.
The algorithm produced a variable matching ratio in which a case may
have one or more matched controls. The results of this process are
summarized in Table 15.
43
-------
DATA ANALYSIS
The Method of Multiple Logistics
Multivariate risk analysis is motivated by the need for methods that
assess the direct predictive strength associated with each member of a
cluster of interrelated risk factors. With univariate analysis, each
measure is examined separately for its relation to disease incidence.
However, only multivariate risk analysis provides a method for determining
whether (l) both measures have direct predictability for disease, or (2) one
appears as a predictor in the univariate sense wholly or in part because of
its association with the other.
The essence of multiple risk analysis is to examine changes of risk
with varying levels of one factor at various fixed levels of the remaining
factors. When many risk factors are being considered, a traditional method
for multiple risk analysis based on cross-tabulation is not feasible.
Erratic results occur because too few cases of disease must be distributed
over too many risk subgroups. Instead, multiple risk analysis can be
approached with use of risk models.
The multiple logistic risk model is the most widely used. It is
constructed so that risk of disease is automatically restricted to the
appropriate range, 0 to 1. In this model risk, R, is represented by the
equation
where BQ, B1 , • ••B,< are the k + 1 logistic coefficients. These are
estimated from data which measure levels for k risk factors
(X1 ,7^2 »• • • »Xj^) for each subject.
The rather formidable logistic model can be simplified somewhat by
linearizing it to give In R/(1 - R) = BQ + B^X1 + B2X2.. . + B^X-K.
The quantity, In R/(l - R), is called the "logit" of risk and is the
mathematical transformation of risk from which this model derives its name.
By thinking in terms of logit of risk, which ranges from minus infinity to
plus infinity a3 risk ranges from 0 to 1, it is possible to interpret the
logistic coefficients, 3^ ,^21...jB^. They measure the change in logit
of risk per unit change in the respective risk factors.
Descriptive Statistics
Tables 16 and 17 present the mean and standard error for a number of
variables used in the analysis. Data are given by sex and case-control
status.
R = 1/
• • • + Bi^X^)
44
-------
TABLE 16. DESCRIPTIVE STATISTICS FOR MALES. MEANS AND STANDARD
DEVIATIONS FOR CASES (N= 144) AND CONTROLS (N-217)
Cases Controls
Standard Standard
Variable Name Mean deviation Mean deviation
Air pollutants
Total suspended
particulate* 1424-30
Lead* 11.49
Sulfate* 176.01
Nitrate* 146.42
Benzene soluble
organics* 95-72
Mutagenicity** 122.96
Mutagenicity with S9** 149.63
Benzo(a)Pyrene*** 3*78
Benz(a)Anthracene*** 2.57
Chrysene*** 3.12
Benzo(ghi)Perylene*** 22.65
Industry, duration (yrs)
Asbestos related 3-50
Petrochemical 1 .57
Construction 2.63
Metal 3.58
Other 12.08
(continued)
942.55 1275.88 803-58
7.58 10.80 6.20
114.33 160.65 99.76
94.49 136.24 77.46
63-59 88.75 50.85
77.95 119-83 75-27
94-47 146.03 89.18
3.64 3.53 2.56
2.78 2.29 1.81
3-12 2.80 2.00
18.00 22.05 14.18
7.79 1.91 5.70
5-62 1-35 5-60
8.10 2.58 6.09
8.09 1.88 6.53
13-78 7.42 11.57
45
-------
TABLE 16 (continued)
Cases Controls
Variable Name
Mean
Standard
deviation
Mean
Standard
deviation
Water source (yes/no)
Mokelumne River
31%
4-%
52!?
San Pablo
2\%
3%
132
2%
Sacramento River Delta
42%
4%
35%
3%
Lifestyle
Smoker (ever/never)
91%
1 ^
1/0
14%
3%
Smoking duration (yrs)
38.10
13.45
24.95
18.58
Smoking (pack yrs)
53-97
35.15
26.19
27.86
Smoking
(Average # packs/day)
1 -34
0.70
0.79
0.70
Alcohol (servings/week)
20.95
34.07
10.61
16.21
Diet
Green vegetables
(servings/week)
5.17
2.42
5-59
2.30
Yellow vegetables
( s e rvi ngs/we ek)
4.28
2.49
4.44
2.46
Demographic
Education (yrs)
12.80
6.00
14.00
3-20
Income****
$17,000
$8,000
$21 ,500
$8,200
*Yr x
**Yr x rev/m^
***Yr x ng/w?
****Cases (N = 11 "5), Controls (N = 184)
46
-------
TABLE 17. DESCRIPTIVE STATISTICS FOR FEMALES. MEANS AND STANDARD
DEVIATIONS FOR CASES (N = 105) AND CONTROLS (N = 156)
Cases Controls
Standard Standard
Variable name Mean deviation Mean deviation
Air pollutants
Total suspended
particulate* 1155-54
Lead* 9*59
Sulfate* 144.85
Nitrate* 120.93
Benzene soluble
organics* 79-86
Mutagenicity** 107.27
Mutagenicity with S9** 130.66
Benzo(a)Pyrene*** 3.19
Benz(a)Anthracene*** 2.17
Chrysene*** 2.52
Benzo(ghi)Perylene*** 19-31
Industry, duration (yrs)
Asbestos related 0.32
Petrochemical 0.17
Construction 0.32
Metal 0-54
Other 1.59
(continued)
879.59 1284.54 977.82
7.33 10.96 8.31
113-51 167.58 150.59
88.92 129.49 94.75
59.98 86.72 64-06
79.53 112.41 83.82
96.53 104.75 104.81
3.01 2.89 2.84
2.27 1.85 2.04
2.27 2.52 2.63
15.79 18.87 15-50
1-67 0.18 0.59
1.5 0.02 0.20
5.50 0.01 0.14
5-11 0.52 1.81
5-81 1.69 5-62
47
-------
TABLE
17 (continued)
Cases
Controls
Variable name
Mean
Standard
deviation
Mean
Standard
deviation
Water source (yes/no)
Mokelumne River
45?
5%
43%
4%
San Pablo
17?
4%
21%
3%
Sacramento River Delta
CO
a*
5%
36$
4%
Lifestyle
Smoker (ever/never)
00
3$
58%
4%
Smoking duration (yrs)
53-11
15.91
19-98
19-27
Smoking (pack yrs)
39-30
27.29
14.27
17-52
Smoking
(Average # packs/day)
1 .03
0.66
0.43
0.48
Alcohol (servings/week)
10.93
23.61
5.67
9.20
Diet
Green vegetables
(servings/week)
5-57
2.71
5.97
2.67
Yellow vegetables
(servings/week)
4.22
2.21
4.27
2.62
Demographic
Education (yrs)
12.76
.96
13.92
2.16
Income****
$13,700
$8,700
$14,900
$7,900
*Yr x yg/m-^
**Yr x rev/m^
***Yr x ng/m^
****Cases (N = 79), Controls (N = 120)
48
-------
Analytic models
Fitting Smoking Variables
Smoking is known to be the single largest cause of lung cancer in the
population. As shown in Table 16 of this report, 97$ of male cases fit the
criterion for being a regular smoker as compared to 74% of the controls.
The comparable figures for females (Table 17) are 86% of cases and 58% of
controls. Consequently, it was important to identify the best measurement
representing smoking so that all other variables could be examined, adjusted
for the effect of 3moking. The '"best" smoking measurement was that smoking
variable or combination of smoking variables which explained the largest
amount of lung cancer.
To account adequately for smoking and its relationship to lung cancer
in this data set, a number of logistic models of lung cancer and smoking
were tested and reviewed.
Available smoking variables were 1) ever or never smoked regularly,
i.e., one or more cigarette per day for 12 months, 2) duration, i.e., total
number of years smoked, 3) average packs per day smoked, and 4) total pack
years, i.e., packs per day times duration smoked. Variables were "fitted"
singly and in pairs for males and females as separate analytic groups.
Interaction terms, as well aa various transformations (e.g., the square of
the variables) were modeled.
A simple "best fit" model was selected. Two measures, smoking duration
and average packs per day used as independent terms in a single model,
proved to be statistically significant for males. For females, only the
average packs per day proved to be statistically significant, but for the
purpose of parallel presentation, the smoking duration variable was also
retained. Therefore, both measures of smoking (duration and average packs
per day) were used in all subsequent analytical models.
Analysis with Additional Variables
Table 18 presents the odds ratios computed via the multiple logistics
analysis technique for two environmental factors, both with and without
controlling for the effect of smoking. None of the odds ratios are
statistically significant.
Table 19 presents the odds ratios for several of the measured air
pollutants in Contra Costa County, controlled for known risk factors. The
dose values pertain to the total cumulated exposure from residence in the
county. Four examples for males are presented. No value for any pollutant
for either males or females approached statistical significance, with or
without controlling for known risk factors.
49
-------
TABLE 18. RISK* OF LUNG CANCER, BY SEX, ASSOCIATED WITH SELECTED
ENVIRONMENTAL FACTORS IN CONTRA COSTA COUNTY, WITH AND WITHOUT
CONTROLLING FOR SMOKING
Males Females
Factor OR p** OR p**
Industrial area
Not controlled
Controlled for smoking
Near waste dumps
Not controlled
Controlled for smoking
1.68 0.06
1.55 0.17
1.04 0.94
0.97 0.95
0.71 0.40
0.53 0.21
0.76 0.75
0.67 0.67
*Risk expressed as the odds ratio (OR)
**Significance probability
TABLE 19- RISK* OF LUNG CANCER FOR VARIOUS AIR POLLUTANTS, CONTROLLED
FOR SMOKING, DRINKING AND ASBESTOS EXPOSURE, CONTRA COSTA COUNTY MALES
1 - ¦ — - 1 ¦ ¦¦¦ ¦' —. .1. ¦ 1 ' ~ — — I I.I I. I ¦ I ¦ ¦ I
Factor
OR
p**
SO^ dose (yg/m^ from
county residence)
1 .002
.155
NO-j dose (yg/m^ from
county residence)
1 .003
.142
Mutagen dose (rev/m^
from county residence)
1 .001
.460
Benzo(a)pyrene (ng/m^
from county residence)
1 .032
• 505
*Risk expressed as the odds ratio (OR)
**Significance probability
50
-------
TA3SE 20. Eto\* 0? SG CANCER FCE TAEIOUS OCCUPATIONAL CATEGORIES,
eOWTSDLlEj SHC«TXGr ERITTKIKS, ASE25TC? KP0SU53 JOT SO, EOSSt
CONTFA COSTA C DOTTY KALES
Fastens
OK
Mstal irjies.tr/ 'cgc yss;- cf ewploy^anl;)
"Other" industries (je™ y&sr ef er.ployaent ]
Construction, industry (per year of employment)
Petrocheadcal in this try (per year af eurpIojireri-O
1 .016
J .009
1.024
0.995
433
AyZ-
• 254
•tLafls ajrsisage-d e_& -fchs ;i1a ratio VOT.1
~Significance probability
Table 23 presents tha odds rstio3 far the broad occupational categories
of aala Hue collar warksr. Wone of the odds ratios approached statistics!
aignificer.ce, with or without controlling for knava risk iactara.
Tables 21 and 22 present the odds ratios for each factor iacltdee i.i
the saturated model for raalea and femalea, respectively. This asultipls
losLatic acaljrsis- uairig 13 vansWes iVi not si:batantislJy slier the adds
rati3 t~r any faetar fr=r: -tat found la simpler nod-sla. Ftrrh-sr, this
=5-c^.Tifc. »i>jal failed te- ^jcalai:! -jet: no:» af ta= Ictg ctn-3-sr tiEtr 313
the simpler models iising onljr variables with either a s'tafciafcicalLp-
significant relationship to lung cancer or s known causal relationship to
luEg CEjicef eatafcliahed free, otliar studies.
51
-------
TABLE 21. RISK* OF LUNG CANCER FOR EACH FACTOR IN THE SATURATED MODEL
FOR MULTIPLE LOGISTIC ANALYSIS, CONTRA COSTA COUNTY MALES
Factor
Unit
OR
p**
Smoking duration
(per year smoked)
1 .039
< .0005
Smoking dose
(per pack per day)
1 .396
.007
Green vegtables
(per weekly serving)
0.809
.012
Alcoholic drinks
(per weekly serving)
1 .008
.237
Asbestos exposure
(per year of exposure)
1 .035
.188
SO^ dose
(total yg/m^ from county)
1 .001
.547
Petrochemical industry
(per year of employment)
0.937
.617
Construction industry-
(per year of employment)
1 .025
cn
CM
Metal industry
(per year of employment)
1 .021
• 354
"Other" industries
(per year of employment)
1 .008
.541
Mokelumne River water
(source at interview)
0.620
.280
Sacramento delta water
(source at interview)
0.724
.488
Yellow vegetables
(per weekly serving)
1 .022
.780
*Risk expressed as the odds
**Significance probability
ratio (OR).
52
-------
TABLE 22. RISK* OF LUNG CANCER FOR EACH FACTOR IN THE SATURATED MODEL
FOR MULTIPLE LOGISTIC ANALYSIS, CONTRA COSTA COUNTY FEMALES
Factor
Unit
OR
p**
Smoking dose
(per pack per day)
10.249
.004
Smoking duration
(per year smoked)
0.998
.910
Green vegetables
(per weekly serving)
1 .194
.176
Alcoholic drinks
(per weekly serving)
0.967
.252
Metal industry
(per year of employment)
1 .210
.244
"Other" industries
(per year of employment)
0.928
• 357
Yellow vegetables
(per weekly serving)
0.897
.442
Petrochemical industry
(per year of employment)
1 .327
.477
Sacramento delta water
(source at interview)
1 .072
• 923
Mokelumne River water
(source at interview)
0.919
• 906
Asbestos exposure
(per year of exposure)
1 .069
.780
Construction industry
(per year of employment)
1 .160
.917
SO^ dose
(total yg/m^ from county)
1 .001
.811
*Risk expressed as the odds ratio (OR).
**Significance probability
53
-------
TABLE 23. THE RISK* OF LUNG CANCER FOR MALES SMOKING THE AVERAGE DOSE
AND DURATION OF SMOKING MALE CASES (3=140) AND F03 FEMALES SMOKIHG THE
AVERAGE JOSE OF SMOKING FEMALE CASES (l!T= 90)
Males Females
38.5
NS**
1 .21
10.25
15.47
*Risk expressed as the odds ratio
**Kot used in the computation since there was ao statistically
significant effect
To assess the approximate contribution to lung cancer risk by smoking,
several methods were employed. The first estimated the effect from the
average daily dose and the average smoking duration among all male cases who
were smokers and for female cases, the average daily dose smoked among
smokers. The results, presented in Table 25, illustrate that the average
effect of the statistically significant smoking variables on the risk for
lung cancer is similar for both males and females, e.g., 11.28 and 15.47
times that of nonsmokers, respectively.
The second nethod computed the proportion of the "explainable" lung
cancer contributed by smoking, under alternate conditions (Tables 24 and
25) (Coles, 1980). There are three alternative conditions for accounting
for lung cancer by the variabiles under study. The first is to require that
only those factors statistically significantly related to lung cancer in the
analysis be considered candidates for a causal relationship. This is the
most conservative condition and, under this condition, the amount of lung
cancer "explained" by the data, and the analytic model, is only that amount
"explained" by those factors bearing a statistically significant relation-
ship to lung cancer. Under this conservative condition, 29$ of all lung
cancer among females and 27$ of all lung cancer in males can be accounted
for by this analysis. Of that part which can be accounted for, 100$ in
females and 33% in males is contributed by smoking.
The second alternative condition is to allow those factors with
statistically significant relationships to lung cancer plus those factors
that have been shown by other studies to have a causal relationship to lung
cancer (even if in this study the relationship does not reach statistical
Average years smoked
39.2
Risk per year
1 .04
Average packs smoked per day
1 .38
Risk per pack
1 .90
Total average smoking risk
11 .28
54
-------
TABLE 24. THE RELATIVE CONTRIBUTION BY EACH RISK FACTOR, UNDER THREE
ASSUMPTIONS*, TO THE EXPLAINABLE PROPORTION OF MALE LUNG CANCER IN
CONTRA COSTA COUNTY, EXPRESSED AS CUMULATIVE PERCENTS
Cumulative percent contributed
by Assumption*
Risk factor p 12 3
Smoking dose
.007
83.3
76.7
70.9
Smoking duration
<•0005
Green vegetables
.012
100.0
92.1
85-1
Alcoholic drinks
.237
-
95-8
88.5
Asbestos exposure
.188
-
100.0
92.3
SO4 dose
• 547
-
-
93.8
Petrochemical industry
.617
-
-
94.1
Construction industry
.293
-
-
95.8
Mokelumne River
.280
Sacramento Delta
.488
Metal industry
• 354
>
-
100.0
Other industries
.541
Yellow vegetables
.780 -
Proportion of all lung
cancer explained: .268 .291 .315
*Assumption 1 . The most conservative assumption: Lung cancer can be
causally related only to those factors to which it is statistically
significantly related (p<0.05).
Assumption 2. The moderate assumption: Lung cancer may be causally
related to those factors to which it is statistically significantly related
(p<0.05) and also to those factors known from other information to be
causally related.
Assumption 3. The most liberal assumption: Lung cancer may be causally
related to all factors studied, irrespective of the statistical
significance of the relationship.
55
-------
TABLE 25. THE RELATIVE CONTRIBUTION BY EACH RISK FACTOR, UNDER THREE
ASSUMPTIONS*, TO THE EXPLAINABLE PROPORTION OF FEMALE LUNG CANCER IN
CONTRA COSTA COUNTY, EXPRESSED AS CUMULATIVE PERCENTS
Risk Factor
P
Cumulative Percent Contributed
by Assumption*
1 2 3
Smoking dose
.004
100.0 1
93.2
81 .1
Smoking duration
• 910
- J
Green vegetables
.176
-
95-5
83-1
Alcoholic drinks
.252
-
99.8
86.9
Asbestos exposure
.780
-
100.0
87.0
SO4 dose
.811
-
-
87.0
Petrochemical industry
.477
-
-
90.5
Construction industry
• 917
-
-
91.0
Mokelumne River
.906
Sacramento Delta
• 923
Metal industry
.244
»
-
100.0
Other industries
• 357
Yellow vegetables
.442
J
Proportion of all lung
cancer explained:
.294
.316
• 363
*Assumption 1 . The most conservative assumption: Lung cancer can be
causally related only to those factors to which it is statistically
significantly related (p<0.05).
Assumption 2. The moderate assumption: Lung cancer may be causally
related to those factors to which it is statistically significantly related
(p<0.05) and also to those factors known from other information to be
causally related.
Assumption 3- The most liberal assumption: -Lung cancer may be causally
related to all factors studied, irrespective of the statistical signifi-
cance of the relationship.
56
-------
significance) to be considered as candidates for causal relationships. This
is a moderate condition and allows the effect of alcohol use and asbestos
exposure to help account for lung cancer, even though in this analysis
neither were found to bear a statistically significant relationship to lung
cancer. Given that asbestos and alcohol consumption are generally accepted
risk factors for lung cancer, the lack of statistical significance may be
due to either a very small contribution by each factor to lung cancer in
this study or that the cases and controls did not differ much in their
asbestos and alcohol exposure. Under this moderate condition, "52% of the
lung cancer among females and 29$ of the lung cancer among males can be
"explained". Of the explainable portion, 93$ for females and 77$ for
males is contributed by smoking.
The third and most liberal condition is to allow all variables to be
considered eligible candidates for accounting for lung cancer, regardless of
the statistical significance of their relationship to lung cancer. Under
this condition a maximum of J>6% among females and 32$ of the lung cancer
among males is explainable by the saturated analysis model. Of this
explainable amount, smoking contributes 81$ among females and 71$ among
males.
While this last condition is clearly unacceptable for the purposes of
concluding causal relationships, it has the benefit of demonstrating the
maximum contribution by those variables studied if they were to have a
causal role. For example, even though SO^ exposure does not have a
statistically significant relationship to lung cancer in this analysis, it
is possible to estimate that only about 1 .5$ of the total explainable male
lung cancer (i.e., less than 1/2$ of all male lung cancer) and none of the
female lung cancer could be attributed to SO^ exposure if it were a causal
factor.
DISCUSSION
In any analysis a major goal is to be able to explain as much of the
disease of interest as possible with the information gathered for
analysis by the study. In this study several variations on this goal
were also sought. One was to determine whether any lung cancer could be
attributed to constituents of air pollution. The other was to explain
the simultaneous occurrence of three factors in Contra Costa County:
higher incidence of lung cancer in males, higher proportions of blue
collar workers and higher levels of SO^ in ambient air. Therefore the
analyses also sought to explain the earlier findings and to reconcile as
many of the facts regarding lung cancer in Contra Costa County as
possible.
It is unrealistic to expect that a study explain, through its gathered
data, all of the cases of a disease with multiple causation. Reasonable
expectations are that observed differences in the rate of disease between
areas or groups be explained and that the information gathered account for a
significant portion of the disease.
57
-------
TABLE 26. COMPARISON OF THE RISK* OF LUNG CANCER, BY SEX, FOR AMBIENT AIR
SULFATE EXPOSURE MEASURED AS THE AVERAGE ANNUAL LEVEL OF THE CENSUS TRACT
RESIDENCE AT DIAGNOSIS (SO4 LEVEL) AND AS THE COMPUTED TOTAL LIFETIME
OF DOSE IN CONTRA COSTA COUNTY (SO/. DOSE)
Males Females
Factor OR p** OR p**
SO4 level 1.26 0.01 0.94 0.78
S04 dose 1.00 0.23 1.00 0.41
*Risk expressed as the odds ratio (OR)
**Significance probability
To determine whether the data collected at the individual level
corroborated the previous findings from this project, several analyses were
conducted that did not neccesarily test causal hypotheses. The earlier
incidence analysis showed that residents of the Industrial area experienced
lung cancer at a rate approximately 40$ higher than residents of the Non-
Industrial area for the period 1975-79- The cases and controls in this
component of the study were categorized as to their residential address at
the time of interview (for controls) or at the time of initial diagnosis
(for cases). Table 18 shows that when male participants are so categorized
there is an elevated odds ratio (p = 0.06) associated with residence in the
Industrial area, as previously found. When smoking is taken into account,
however this relationship is essentially destroyed (p = 0.17), indicating
that in this set of data, the difference in lung cancer risk for males
between Industrial and Non-industrial areas can be substantially accounted
for by a difference in smoking practices of the males of the two areas. For
females no elevated odds ratio for residence in the Industrial area exists.
Proximity to known locations of toxic waste dumps had no effect on lung
cancer risk, as shown in Table 18.
The previous correlation analysis produced an association between
ambient SO^, percent blue collar workers and lung cancer among males.
That analysis used, as a measure of SO^, the level assigned to the census
tract at the time of diagnosis. Table 26 shows that for males, a positive
but not statistically significant relationship exists between the level of
SO^ in the census tract at the time of diagnosis (for cases) or interview
(for controls). If ambient SO^ exposure really imparts an increased risk
of lung cancer one would expect a more precise measure, estimated cumulative
dose of ambient SO^ in the county, to demonstrate a stronger relation-
ship. However, when the total cumulated dose of ambient SO^ in Contra
Costa County is considered, no relationship exists between SO^ and lung
cancer. Adjusting for smoking or other factors, whether significant or not,
58
-------
did not materially change this finding.
This analysis examined the effect of smoking and air pollution in
considerable detail. No measure of air pollution was found to have a
statistically significant relationship to lung cancer. In considering these
negative findings, the following must be weighed.
1. The estimated cumulative dose of each pollutant was computed from
values in Contra Costa County only. If the population in Contra
Costa County were highly mobile it could dilute the contribution
by each respondent such that a weak relationship could not be
detected. In this situation, however, no correlation with the
pollutant level at the place of residence at diagnosis would
represent a causal relationship unless the relationship were both
immediate and transitory, a requirement not met by any other known
carcinogenic agents.
2. Air pollutants were measured for only one year. Thus if that year
were atypical for the relative levels of each pollutant from area
to area or if measurement of levels were not typical of the census
tract, a real relationship would be obscured through
misclassification.
3. Another assumption is that each subject is locationally fixed at
his/her census tract of residence; consequently this excludes any
other locational exposure such as those in the workplace.
4. Only solid air pollutants were measured. At the start of the
project no reliable method existed for measuring the mutagenicity
of volatile gases. Also, the project was not funded to measure
levels of volatile gases, such as benzene, which could have
biologic effects at very low doses. However, in general the
volatile and particulate pollutants originate from the same
emission sources and the general air flow patterns remain the same
from year to year so that if there were a significant carcinogenic
effect from a volatile pollutant that was not measured, it would
be expected to exhibit its effect in a distribution similar to
that of a particulate pollutant from the same source.
Even with these caveats, the failure to detect any effect on lung
cancer by air pollution suggests that no significant effect exists.
The proportion of lung cancer explainable by this analysis was only
about a third, most of which was contributed by smoking. This may be due to
at least three reasons. The first is that the multiple logistics analysis
draws its information about risk from differences between cases and
controls. With regards to smoking, a high proportion of both cases and
controls smoked. This may have led to an underestimate of the risk from
smoking.
The occupational categories are very broad and undoubtedly contain
specific occupations that are of higher and lower risk than the mixture that
59
-------
constitutes the broad category. The occupational analysis therefore likely
explains less lung cancer than potentially it could. This would be expected
to be more true of males than of females since in general, males would be
expected to have 2 higher proportion of their numbers in occupations with
carcinogenic hazards. This supposition is supported by the fact that higher
proportion of lung cancer among females is explained, under any assumption,
than among males.
Lastly, a major determinant for any cancer was not measurable by this
study. That factor is individual susceptibility. It may explain a
significant proportion of lung cancer but is not identifiable by a
comparison of exposures of cases and controls.
SUMMARY Am CONCLUSIONS
This analysis of case-control data suggests that the major contributor
to lung cancer in Contra Costa County la smoking. Further, smoking accounts
for most of the previously identified difference in lung cancer incidence
between the Industrial and Non-industrial areas. Because of the high
prevalence among both cases and controls, the contribution of smoking may be
underestimated.
There was no identified effect on lung cancer risk contributed from any
measured constituent of air pollution. The one air pollutant (SO^)
significantly correlated with male lung cancer incidence in the indirect
correlational -analysis, in this case-control analysis had a positive but not
statistically significant relationship with lung cancer risk only when 50^
level at the current address was used as the measurement. When a measure of
total lifetime dose of S0^ from Contra Costa County was used, no elevated
risk was apparent.
One dietary factor had a significant (p = 0.01) protective effect for
males and a similar but not statistically significant (p = 0.18) effect for
females. This factor, weekly servings of green vegetables, is a crude
measure for several dietary constituents believed to reduce the risk of
cancer of several types. Both vitamin A and cruciferous vegetables would be
included in this dietary measure. The dietary measure, weekly servings of
yellow vegetables, did not discriminate between cases and controls.
None of the broad occupational categories had any significant
relationship to lung cancer risk in males. A more detailed analysis of the
effect of various occupations on lung cancer risk is desirable and support
for this subsequent analysis is being sought.
The effects of alcohol and asbestos exposure, as measured, did not bear
a statistically significant relationship to lung cancer in this analysis.
In any subsequent analysis a more quantitative measure of asbestos exposure
would be desirable.
There was no apparent effect of source of drinking water or proximity
to known toxic waste dumps on the risk of lung cancer.
60
-------
These data confirm the known causal relationship between smoking and
lung cancer. They provide some reassurance that constituents of particulate
air pollution do not contribute measurably to the risk of lung cancer. This
is consistent with the findings of several other studies. These data
provide supportive evidence for the protective effect of dietary factors on
cancer risk, a finding consistent with other epidemiologic and laboratory
studies.
61
-------
ElgllC-SP.ABjff
AIHZ. zf Tcts3 Cr.gaTi5.r; ffsteriaig in At"ospivsric
Particulate Matfcsr- £IFL 3 3c. 6~ , Tslifcrrjo. Stats Ec-.F-~i.aieT';
ct" Wealth Services, Eerkeiay, California, " S'7?-
MHj. Th-e Ctetiical snA Biochemical Characterisation of Particulate
Matter as Part of ail Epiceniologics! C^iscar Stady. A7>II JTe-t^oi Uo. S1 .
California State Cepartoen:: of r.^altb Services, Berkeley, California,
igeoa.
AIKL. Determination of Sitrates in Atwscur.c Partisii-'jte S!?Hf
(Erucir^e Method). UHI. fethacl TTe. 66, Caiifcmie Stits Vep&of
y.eaVth Services,. Sesfcelar, ¦Halifomi.a, lgSCt.
isTHlo. XieterminatiBrj of S^lfat^ High Voiujce Particulata Samples:
^'~rciJVEa-ric Bariuz Sulfite AI?i Xsttod Ho, 61 . California
State re-partusat of Health Services, Berkeley, California, tgSOc.
Ames, B-, J. MeCsail and S. Yassasaki* Ksthod for Detecting C&rei:tog-3as
and Mutagens witfe the Salmon«lla't*-iainmali£naicro3ome i-'utegeoioi~y Cast.
fetation Bes. , 315347-364, f975-
EAADHD. Total Su&pendsi Psrticulats Gravimetric Analysis Procedure. Bay-
Area Air Sualxtj Manager.snt "District, San "Francisco, Califo-cnia, 19TT.
BeEmia, J- A. Stain acS J- L. Andrews, Jr. Changing Bpidereiclogy of Lueg
Cancer. Increasing ^ncideacs ir. Women. ^a3. Clia. S- Aa., 59(2,Js3'5-
32-i, JS?5-
31ofc, «. J.r !¦ A. "Briton, Stone. Cancer
Ko-tality in fJ.S- Counties virh ?strol&uE Industries. Science, 198,
51-5?, /=>??.
Elotf V. J, ?. Jrausstii, Jr. Arasricsl Atr Folia{¦£«» a-i-d lung Cancer,
?>.a jjais:, July 26, "9"5.
Butfll, P. snc J. 3. Jr. Relative uraast of Smoking and Air Pollution
on Lung Cancer. Areru Ito** >5:39'-23"1» 1967.
Burbank, J. U.S. Lass ^tr-csr l>eata Hs~.»s- 1-egi.ri tc f.ise Praportxc-r^sfeJy Vis re
Rapidly for- Jamaica Tliaji for Kales; A Dose fie-spcmae ^flectT J.
Ciron, Dis. 25;i13-4~9, '975.
62
-------
Bureau of the Census. 1980 Alphabetical Index of Industries and
Occupations. U.S. Department of Commerce, Washington, D.C., 1981.
267 pp.
Carnow, B. W. The "Urban Factor" and Lung Cancer: Cigarette Smoking or Air
Pollution? Environmental Health Perspectives. 22:17-21, 1978.
Clemmesen, J. Registration of Data Concerning Cancers Possibly
Produced by Air Pollution, In: Air Pollution and Cancer in Men,
U. Mohr, and T. L. Schmahl, eds. IARC Scientific Publ. No. 16.
International Agency for Research on Cancer, Lyon, 1977.
Coles, et al. Determining the Most Valuable Clinical Models. Methods
Information In Medicine, 19(l):41-49> 1980.
del Regato, J. A. and H. J. Spjut. Ackerman and del Regato's Cancer
Diagnosis and Prognosis. The C.V. Mosby Company, St. Louis, Missouri,
1977.
Doll, R. and R. Peto. The Causes of Cancer: Quantitative Estimates of
Avoidable Risks of Cancer in the United States Today. JNCI 66(6):1 191 —
1308, 1981.
Flessel, C., J. J. Wesolowski, et al. Integration of the Ames Bioassay
and Chemical Analyses in an Epidemiological Cancer Incidence Study,
pp. 67-8^. In: Application of Short-Term Biossays in the
Fractionation and Analysis of Complex Enviromental Mixtures,
M. D. Water, S. Sandhu, et al, eds. Vol. 2. Plewum Publishing
Corporation, New York, 1981 .
Fraumeni Jr., J. F. Persons at High Risk of Cancer. An approach to Cancer
Etiology and Control. Academic Press, New York, 1975.
Friberg, L. and R. Cederlof. Late Effects of Air Pollution with Special
Reference to Lung Cancer. Environmental Health Perspectives, 22:45-66,
1978.
Gordon, R. J. Personal Communication, 1978.
Hammond, E. C. Smoking Habits and Air Pollution in Relation to Lung
Cancer. In: Environmental Factors in Respiratory Disease,
D. H. K. Lee, ed. Academic Press, New York, 1972.
Hammond, E. C. Tobacco, In: Persons at High Risk of Cancer. An Approach
to Cancer Etiology and Control. J. F. Fraumeni, Jr., ed. Academic
Press, New York, 1975-
Harvard University Laboratory for Computer Graphics and Spatial Analysis.
SYMAP. Version 5.2, Harvard University, Cambridge, Massachusetts,
1975.
63
-------
Henderson, B. E., R. J. Gordon, H. R. Menck, et al. Lung Cancer and Air
Pollution in South Central Los Angeles County. Am. J. Epid. 101(6):477—
488, 1975.
Higgins, I. T. Epidemiological Evidence on the Carcinogenic risk of Air
Pollution, In: Air Pollution and Cancer in Man. U. Mohr,
T. L. Schmahl, eds. IARC Scientific Publ. No. 16. International
Agency for Research on Cancer, Lyon, 1977.
Linden, G. and D. F. Austin. A Rapid Reporting Cancer Incidence System.
American Journal of Epidemiology. 99(3). The John Hopkins University,
1974.
Lombard, H. L. and E. P. Huyck. An Epidemiological Study of Lung Cancer
Among Females. Growth, 52:41-56, 1968.
Moore, H. Application of Wavelength Dispersive X-Ray Fluorescence
Spectrometry to the Determination of Lead in Atmospheric Particulate
Matter Collected on High-Volume Glass Fiber Filters. AIHL Report 183.
Air and Industrial Hygiene Laboratory, California Department of Health
Services, Berkeley, California, 1976.
NCI. Third National Cancer Survey: Incidence Data. National Cancer
Institute Monograph 41. Publication No. (NIH) 75-787, NCI, NIH, Public
Health Service, U.S. Department of Health, Education and Welfare.
NCI. Incidence and Mortality Data, 1973-1977. National Cancer Institute
Monograph 57. Publication No. (NIH) 81-2330. NCI, NIH, Public Health
Service, U.S. Department of Health, Education and Welfare, 1981.
Oliphant, T. H. and R. B. McHugh. Least Significant Relative Risk Determi-
nation in the Case of Unequal Sample Sizes. American Journal of
Epidemiolgy. 113(6). The John Hopkins University, 1981.
Pike, M. C., R. J. Gordon, et al. Air Pollution, In: Persons at High
Risk of Cancer. An Approach to Cancer Etiology and Control.
J. F. Fraumeni, Jr., ed. Academic Press, New York, 1975.
Pike, M. C., J. S. Jung, et al. In: Energy and Health. N. E. Breslow and
A. S. Whittemore, eds. pp. 3-16. SI AH Institute for Mathematics and
Society, Philadelphia, 1979•
Vena, J. E. Air Polution as a Risk Factor in Lung Cancer. Am. J. Epid.
116(0:42-56, 1982.
Williams, R. R. and J. W. Horn. Association of Cancer Sites with Tobacco
and Alcohol Consumption and Socioeconomic Status of Patients:
Interview Study from the Third National Cancer Survey. J. Natl.
Cancer Inst. 58(3):525—547, 1977.
64
-------
World Health Organization. International Classification of Diseases
for Oncology. World Health Organization, Geneva, Switzerland, 1976.
Young, Jr., J. P., C. H. Percy, A. J. Asire, eds. Surveillance
Epidemiology and End Results: Incidence and Mortality Data. National
Cancer Institute Monograph 57. NIH Publ. 81-2330, National Cancer
Institute, Bethesda, Maryland, 1981.
65
------- |