This article was downloaded by: [University of Alberta] On: 30 December 2014, At: 19:16 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Archives of Environmental Health: An International Journal Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/vzeh20
Ambient Air Pollution and Cancer in California Seventh-day Adventists a
a
a
Paul K. Mills Ph.D., M.P.H. , David Abbey Ph.D. , W. Lawrence Beeson M.S.P.H. & Floyd Petersen M.P.H.
a
a
Department of Public Health and Preventive Medicine , School of Medicine, Loma Linda University , Loma Linda, California, USA Published online: 03 Aug 2010.
To cite this article: Paul K. Mills Ph.D., M.P.H. , David Abbey Ph.D. , W. Lawrence Beeson M.S.P.H. & Floyd Petersen M.P.H. (1991) Ambient Air Pollution and Cancer in California Seventh-day Adventists, Archives of Environmental Health: An International Journal, 46:5, 271-280, DOI: 10.1080/00039896.1991.9934387 To link to this article: http://dx.doi.org/10.1080/00039896.1991.9934387
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
Ambient Air Pollution and Cancer in California Seventh-day Adventists
Downloaded by [University of Alberta] at 19:16 30 December 2014
PAUL
K. MILLS,
Ph.D.,
M.P.H.
DAVID ABBEY, Ph.D. W. LAWRENCE BEESON, M.S.P.H. FLOYD PETERSEN, M.P.H. Department of Public Health and Preventive Medicine School of Medicine Loma Linda University Lorna Linda, California
ABSTRACT. Cancer incidence and mortality in a cohort of 6 O00 Seventh-day Adventist nonsmokers who were residents of California were monitored for a b y period, and relationships with long-term ambient concentrations of total suspended particulates (TSPs) and ozone (0,)were studied. Ambient concentrations were expressed as mean concentrations and exceedance frequencies, which are the number of hours during which concentrations exceeded specified cutoffs (e.g., federal and California air quality standards). Risk of malignant neoplasms in females increased concurrently with exceedance frequencies for all TSP cutoffs, except the lowest, and these increased risks were highly statistically significant. An increased risk of respiratory cancers was associated with only one cutoff of 4, and this result was of borderline significance. These results are presented in the context of setting standards for these two air pollutants.
cross-sectional or correlational d e ~ i g n . ~In, ~ this apADVERSE HEALTH EFFECTS associated with ambient proach, indices of air pollution in certain geographic air pollution have been scrutinized by environmental locales were correlated with the age-adjusted cancer scientists for most of the twentieth century.',2 There is mortality rates in the same areas, with or without adevidence that air pollution contributes to morbidity from airway obstructive disease and other forms of resjustment for consumption of cigarettes. Most of these piratory disease, cancer, and cardiovascular di~ease.~.~ studies have not demonstrated that an increased cancer risk occurs with an increase in air pollution levels. It is difficult to evaluate this evidence because the efIn fact, the U.S. Environmental Protection Agency crifects of tobacco smoke are difficult to separate from the effects of air pollution, especially in urban areas where teria document for particulate matter and sulfur oxides the numbers of smokers and concentrations of air polconcluded: ". . . nor does there presently exist credible lutants are higher than in more rural areas. Cigarette epidemiological evidence linking increased cancer smoking has been implicated in the etiology of cardiorates to elevations in particulate matter as a class, i.e., undifferentiated as to chemical content."' In many invascular disease, respiratory diseases, lung cancer, and several other forms of cancer in humans (e.g., cancers stances, however, an inability to detect relationships can be linked with limitations in study design, poor of the bladder and pancreas). Most studies of air pollution and cancer in humans statistical power, or other methodological weaknesses. Seventh-day Adventists provide a unique opportunity have been epidemiologic investigations that used SeptemberlWober1991 [Vol. 46 (NO.5)]
271
for investigating the health effects of ambient air pollutants with very little confounding by tobacco smoke. Use of tobacco or alcohol is proscribed by the Seventh-day Adventist church. However, Seventh-day Adventists‘ contact with ambient air pollution varies greatly by virtue of their choice of residence and occupation. This variability in pollution exposure makes possible the evaluation of the relationship between exposure and health effects, and there is minimal confounding by tobacco smoke. In this report, we present the risks of cancer associated with long-term exposure to ambient concentrations of total suspended particulates (TSP) and ozone (03), which were observed in a cohort study of California Seventh-day Adventists who did not smoke.
Downloaded by [University of Alberta] at 19:16 30 December 2014
Methods The study design and measurement of ambient air pollution have been described in a previous study.’ Enrollment of the study population. In April of 1977, a subgroup of 6 340 members of the National Cancer Institute-funded Adventist Health Study were enrolled in a prospective epidemiological study for the primary purpose of studying the health effects of long-term cumulative contact with air pollution. Persons who reported that they were current smokers were excluded from the study. Because many individuals had not been Adventists all their lives, they may have smoked in the past or may have resided with a non-Adventist who smoked. Therefore, detailed smoking histories were ascertained for each participant. Individuals who met the following criteria were included in the study: (1) age 25 y+ at the time the Adventist Health Study questionnaire was completed in 1974; (2) member of the Seventh-day Adventist church at the time of enrollment in the Adventist Health Study; (3) Non-Hispanic white (individuals of all races completed the questionnaire, but only non-Hispanic whites were followed for cancer and heart disease surveillance); (4) lived at least 10 y within 5 mi of their current residence; and (5) resided in one of three metropolitan areas (San Francisco, Los Angeles and eastward [South Coast Air Basin], or San Diego), or were included in a random sample of 862 individuals who resided in the remainder of California. These 862 individuals represented a 10% random sample of Adventist Health Study participants who met all of the above criteria, except for number 5. Individuals who met these criteria received, by mail, a baseline questionnaire that ascertained their residence history by month and zip code since 1960 and their lifestyle habits pertinent to relative air pollution exposure, e g , work location, hours driving on crowded freeways, percentage of time spent indoorsloutdoors, etc. They also completed the National Heart, Lung, and Blood Institute’s (NHLBl’s) respiratory symptoms questionnaire, which was used to determine selfreported symptoms of chronic respiratory disease. Detailed smoking histories were also obtained as were histories of ever having lived or worked with a smoker and the duration of those exposures. The response fraction to the baseline questionnaire was 87%. A wide range of 272
other lifestyle data (e.g., demographic and dietary information) were already available for these individuals because they were included in the Adventist Health Study database. Cancer monitoring program
Incidence. A surveillance system consisted of annual mailings to every member of the cohort between April 1, 1977, and December 31, 1982. Information on any hospitalization that may have occurred during the prior 12-mo period of follow-up was obtained. If a hospitalization was reported, the name and address of the respective hospital were recorded, and permission to review the resulting medical record was obtained. Adventist Health Study personnel reviewed all medical records for a diagnosis of cancer or cardiovascular disease. Pertinent portions of the medical records were microfilmed to allow a confirmation of the diagnosis by senior medical personnel. This type of follow-up was complete for 99% of the cohort. Computerized record linkage was conducted in areas of California where there were population-based tumor registries (i.e., Cancer Surveillance Program, Los Angeles County; the Resource for Cancer Epidemiology, San Francisco Bay area).g Mortality. Three mechanisms were used to monitor mortality in the study population: (1) computerized record linkage with the California death certificate file, (2) computerized record linkage with the National Death Index, and (3) manual linkage with Seventh-day Adventist church records. These mechanisms enabled us to identify deaths that occurred in the study population during the period of follow-up (1977-1986). Statistical methods Adjusted relative risks (ratios of incidence rates) are presented for average annual exposures to TSP and O3 that exceeded cutpoints versus lower exposures. Relative risks were adjusted for relevant covariates, including age, sex, and past smoking history. Mantel-Haenszel analyses were performed for each outcome before the more sophisticated analyses with proportional hazards regression models. These analyses categorized continuous variables, and any assumptions of linear or additive effects were avoided. The disadvantages of these analyses are the loss of statistical power resulting from this categorization and their inability to adjust for more than a few covariates (two covariates, with the sample size of the present study). The results of the Mantel-Haenszel analysis should not be considered conclusive, but should be used only as a check on the multivariate models. The Mantel-Haenszel analyses were age-adjusted according to the following strata: 25-44 y, 45-64 y, 65-79 y, and 80 y and older. Cumulative annual average concentrations and their annual average exceedance frequencies above several cutoffs were used as indices for long-term ambient concentrations of TSP and 03.Two cutpoints for each index of exposure were determined from the cumulative distributions of the indexes for the Archives of EnvironmentalHealth
Downloaded by [University of Alberta] at 19:16 30 December 2014
subcohorts who resided inside and outside the South Coast Air Basin of California (SCAB). Not much overlap existed between the distributions of the generally lower ambient concentrations outside the SCAB and ambient concentrations inside the SCAB. The lower of the two cutpoints was the 90th percentile of ambient concentrations outside the SCAB, and the higher cutpoint was the median of all inside SCAB concentrations that were above the lower cutpoint. Cox proportional hazards regression models for cancer incidence and mortality outcomes, which controlled for several covariates, were fitted to the data. Cox proportional hazards regression models” are the model of choice for cancer incidence and mortality outcomes when time to event in the cohort is the variable of primary interest. These models controlled for several covariates and were fitted by the BMDP2L stepwise procedure.” Cumulative ambient concentrations of TSP and 0, were represented by annual average concentrations and by annual average exceedance frequencies that were above cutoffs for two time periods (January 1966 to March 1977 and January 1973 to March 1977). The second time period was allowed as an alternative for the longer time periods in the models. We thought that ambient pollutant monitoring was more representative of that experienced in the locations of study participants during this later period because of an increased number of monitoring stations. Because all study participants had lived in their neighborhoods for at least 10 y prior to 1977, we thought the later time period might provide a better ranking of the relative ambient concentrations that were experienced by study subjects for the longer time period. Cumulative ambient concentrations for the two time periods were allowed to compete for entry in the stepwise selection process that was used for original model formulation. Separate models were fitted for annual average concentrations and for exceedance frequencies that were above each cutoff. Time-dependent Cox regression analyses were also performed; average pollutant exposures between January 1973 and the time of risk set as the exposure variables were used. Models that involved ambient concentrations of TSP and 0, were fitted separately. Models were first fit for TSP using exceedance frequencies for the cutoff of 200 pg/m3 (TSP200) and ozone using exceedance frequencies for the cutoff of 10 pphm (0,lO). For each pollutant the variables selected for these cutoffs were then used for the cutoffs as well as for mean concentration. The primary candidate independent variables-total years of smoking, gender, and education-were forced into each model. Age was not among the covariates because age, instead of time in study, was used as the time variable in the models (as recommended by Breslow”). The use of age as the time variable provided a tighter control and adjustment for age. This was especially important for disease outcomes that were strongly age-related. The secondary candidate independent variables that were considered for entry into the models were years lived with a smoker, years worked with a smoker, and past or present employment in an occupation during which high exposures to airborne contaminants occurred. A list of the occupa!kptember/October1991 [Vol. 46 (No.S)]
tions in which the study participants experienced high exposures is given in the study by Euler et aL3 The pollutant exposure variable or variables and the secondary candidate variables competed for entry by the stepwise selection procedure. An F to remove of 0.15 and an F to enter of 0.10 were used to terminate the stepwise procedure. The final models chosen required an F to enter of 0.05 or less. Interactions between TSP or O3 and the other covariates were assessed; first-order cross-product terms in the model were included and then removed, and the change in the log-likelihood was noted.
Results Three different disease outcomes were analyzed in this analysis: (1) all malignant neoplasms for males, (2) all malignant neoplasms for females, and (3) respiratory cancers in all subjects. All malignant neoplasms in males and females were analyzed separately because some cancer sites are sex specific. The distributions of the relevant covariates (e.g., past history of smoking, education) are shown in Table 1. Approximately 62% of the study population resided within the South Coast Air Basin. The frequency distributions of the air pollution variables are shown in Table 2. Between the initiation of follow-up on April 1, 1977, and the end of incidence follow-up on Decernber 31, 1982, there were 301 newly diagnosed cancers (ICDO 140-200) detected in the population (Table 3). Only 17 respiratory cancers were diagnosed. Several demographic and smoking characteristics of these respiratory cancer cases are presented in Table 4. During the mortality follow-up period (1977-19861, there were 180 cancer deaths.
Cancer incidence Results of the stepwise regressions indicated that exceedance frequencies during January 1973 through March 1977 were most closely associated with cancer risk between 1977 and 1982. For this reason, average annual exceedance frequencies during this time period were use in the regression analyses. However, exceedance frequencies for the longer time period (1966-1977) were used to check all results, and only minor differences in the results were noted. The Cox proportional hazards regression models, which were fitted, each included exceedance frequency variables for a cutoff of TSP or 0, for covariates that included gender, education, total years of past smoking, and past or present ernployment in occupations that involved exposure to airborne contaminants. (Occupational exposures were included only for males because females lacked sufficient exposure.) Passive smoking variables did not contribute sufficiently to the fit of the model to merit inclusion. The result of the multivariate analysis of TSP2OO is shown in Table 5. For 1 OOO h/y in excess of this cutoff, the relative risk for all malignant neoplasms among females was 1.37 and was statistically significant fp < .05). Although the relative risk for malignant neoplasms in females associated with mean concentration of TSP was elevated, this was not statistically significant. An 273
examination of the organ site-specific cancer sites that contributed to this excess is presented in Table 6 and indicates that risk of breast cancer was elevated (relative risk [RR] = 1.511, as was risk in all other cancer sites combined (RR = 1.65). Neither of these risk estimates was statistically significant. The relative risk for respiratory cancers for 1 O00 h in excess of TSP200 was 1.72 (Table 7); however, this risk is not statistically significant. Again, the risk associated with mean concen-
Table 1.-Frequency Distributions of Past and Passive Smoking Variables and Covariates in the AHSMOG Incidence Cohort
Downloaded by [University of Alberta] at 19:16 30 December 2014
Variable History of smoking Never Past only
Females
Males
(n = 4 063) NO. %
(n = 2 277) NO.
%
3 504 559
86 14
1 450
827
64 36
3 475 335 196 57
86 08 05 01
1484 339 390 64
675 15 17 03
3 519 292 246 6
87 07
1 507 312 429 30
66 14 19 01
698 1 863 1 149 353
17 46 28 09
358 1170 599 150
16 51 26 07
Education Some high school High school graduate Some college College graduate Unknown
868 695 1715 760 25
21 17 42 19 01
41 7 249 743 856 12
18 11 33 38 01
Occupational air pollution No Yes
4 026 37
99 01
1 964
86 14
2 119 403 1 541 0
52 10 38 0
1 508 236 533 0
66 10 23 0
2 512 815 736
62 20 18
1183 392 702
52 17 31
0
0
0
0
Pack years of cigarette smoking None
< 10
B 10 Unknown Total years of smoking Never 1-9 y
B 1oy Unknown Age (as of 4/1/77) 25-44 y 45-64 y 65-80 y
Years lived with a smoker Never 1-9 y
2 l0y Unknown Years worked with a smoker Never 1-9 y
B l0y Unknown Residence in South Coast Air Basin* Outside Inside Mixed
06
01
Days
Hours" x 1000
313
RR =1.4'
\
TSP THRESHOLD (ug/m3)
* 2 403 3 914 384
32 62 06
*Residence inside or outside the South Coast Air Basin, 1973-1977. Numbers include males and females.
274
tration of TSP was elevated but was not statistically significant. Relative risk contour plots for the associations between all malignant neoplasms in females and TSP exposures are shown in Figure 1. The relative risks for all cutoffs, except for TSP60 and TSP75, are statistically significant. The "dose-response"-type curves for the various cutoffs of TSP are displayed in Figure 2. The results of the multivariate TSP analyses for the other cutoffs and for mean concentration are presented in Table 8. Separate Cox regressions were used for each cutoff and were used with the same covariates as for TSP200. Statistically significant increases in cancer risk were seen among females for all cutoffs, except for TSP6O and TSP75. The magnitude of the increased risk for females is approximately 40-60%. Twofold increases in risk were observed for respiratory cancers, but these risks were not statistically significant. Again there was no relationship observed between TSP ambient concentrations and cancer incidence in the males. The results of the Cox regressions that relate respiratory cancer incidence and O3 ambient concentrations are presented in Table 9. For the 500 h in excess of 0310, the relative risk for respiratory cancer was 2.25, which is of borderline statistical significance. Statistical significance was not achieved for any of the other cutoffs or for mean concentration. The results of the Cox proportional hazards regression analysis were very sim-
**
Risks for TSP60 and TSWS not statistically significant. Average annual hours 1973 - 1977.
Fig. 1. Relative risk contour plots for all malignant neoplasms (female) associated with TSP. Contour plots show incremental increases in hours (days) in excess of different cutoffs required to p m duce stated relative risk.
Archives of Environmental Health
Table 2.-Frequency
Distribution of Total Suspended Particulates and Ozone in the AHSMOG Incidence Cohort
Downloaded by [University of Alberta] at 19:16 30 December 2014
Average annual exposure*
No.
YO
Average annual exposure
TSP60 (hly) Unknown (6000 6 OO1-7 500 >7500
204 1 986 2 429 1721
03 31 38 27
TSP75 (hly) Unknown 6 725
204 2 078 1917 1 743
TSPl00 (hly) Unknown (3000 3 001-5 000 >5000 TSP150 (hly) Unknown < 750 751-2 000 >2000 TSP200 (hly) Unknown < 125 126-1 000 >lo00 TSP mean concentration hg/m3) < 92.41 92.41-114.37 > 114.37
No.
010
0310 (hly) Unknown 600
37 2 216 2 291 1 776
01 35 36 28
03 35 32 29
0312 (hly) Unknown < 25 25-400 >400
37 2 132 2 209 3 745
01 26 27 46
204 2 259 2 288 1 589
03 36 36 25
0315 (hly) Unknown 200
37 2 260 2 305 1 735
01 36 36 27
204 2 237 2 184 1715
03 35 35 27
0320 (h/y) Unknown 50
37 2 260 2 074 1 966
01 35 33 31
204 1 996 2 801 1 339
03 32
44 21
37 2 260 1 585 2 455
01 36 25 38
2 119 1956 2 028
35 32 33
2 779 1 824 1455
46 30 24
0325 (Illy) Unknown 0 %lo > 10 0, mean concentration (pphm) < 2.5634 2.5634-3.1000 > 3.1000
*Average annual hours of exposure to indicated thresholds of total suspended particulates (e.g., 60 or more pg/m3) 1973 to earliest of date of censoring, date of death, or December 31, 1982.
Table 4.-Demographic and Smoking-Related Attributes of Respiratory Cancer Cases in the AHSMOG Incidence Population, 1977-1982
Table J.-Sex-Specific Distribution of Organ Sitespecific Cancer in the AHSMOG Incidence Population Females
Males
Site Colon Rectum Pancreas Lung Leukemia Skin Breast Cervix Uterine Ovary Thyroid Bladder Lymphoma Prostate Total
22 12 8 5 6 9 65 5 37 10 4 0 6 0 189
12 6 4 3 3 5 34 3 20 5 2 0 3 0 100
17 3 0 12 6 16 0 0 0 0 0 7 9 42 112
15 3 0 11 5 14 0 0 0 0 0 6 8 38 100
ilar to the results obtained in the Mantel-Haenszel stratified analysis. Only age and sex were controlled in the stratified analysis. Multipollutant analysis. The relative contributions of September/October 1991 [Vol. 46 (No. 5)]
Attribute
n
Sex Male Female
12 5
Age (y)
(64 65-74 75-84 > 85 Education (8 9-1 2 < 16 Cancer site Female Bronchus Lung Male Lung Larynx Pleura
4 4 8 1 3 5 3
1
Attribute
n
Total years smoked Never 10 < 10 2 > 10 5 Years lived with smokers Never 9 < 10 1 > 10 7 Years worked with smokers Never 10 6 10 1 > 10 6 Hazardous occupation No 15 Yes 2
4
8 2 2
275
Table 5.-cox Proportional Hazards Regression for All Malignant Neoplasms among Females, 1977-1982, with Hours Average Annual Concentration in Excess of ZOO pg/m3 of Total Suspended Particulates USPs) as the Air Pollution Exposure Variable (n = 4 063, Cases = 175)
I
Variable
Regression coefficient
Increment *
Relative riskt
95% C.I. for relative risk
TSP (hours in excess
0.00031745
1 000h/y
1.37
1.05, 1.80
of 200 pg/mW Total years smoked
0.0223
1OY
1.25
0.98, 1.59
Education
0.0424
4Y
1.18
0.94, 1.50
Downloaded by [University of Alberta] at 19:16 30 December 2014
'Increment for computations of relative risk. tRelative risk of increase in exposure of one increment, holding other variables in model constant. *Average annual hours in excess of 200 pg/m3, 1973-1977. §p < .05.
~~
~
~~
Table 6.-Cox Proportional Hazards Regression for Site Specific Cancers among Females Associated with TSPZOO in Excess of 1 OOO h/y (1973-1977) in the AHSMOG Incidence Population ~~
Organ site Breast Cervix Colorectal Respiratoy Leukemiallymphoma All other cancers
Relative risk (95% C.I.)
p Value
1.51 (0.92-2.47) 0.64 (0.05-7.55) 1.08 (0.55-2.1 1) 0.72 (0.12-4.50) 1.05 (0.33-3.37) 1.65 (0.90-3.01)
.10 .72 .83
.72 .94 .10
1.8 -
5000 hrslyr
&
1.6 1.4
1.2
1
TSP THRESHOLD (ug/m3)
*
Risks for TSP60 and TSP75 not statistically significant.
Fig. 2. Relative risk of all malignant neoplasms among females for differing annual average hours (1973-1977) in excess of 60, 75, 100, 150, and 200 pg/m3 of total suspended particulates.
276
TSP and 0, to cancer risk were evaluated. We compared the contributions of the TSP and O3 cutoffs most strongly associated with risk in both the MantelHaenszel and the Cox regression analyses, i.e., 200 pg/m3 and 10 pphm, respectively. With respect to respiratory cancer, 0310 entered the model with a p-toenter value, which was of borderline statistical significance (p = .055), and TSP200 did not enter. Conversely, for all malignant neoplasms among females, TSP 200 entered the model (p to enter = .038), but 0 , l O did not. Cancer mortality. The same Cox proportional hazards regression models were fit to the mortality data as were fit to the incidence data. Covariates included total years of smoking and education (as a surrogate measure of social class). Past or present experience in an occupation with high levels of airborne contaminants was also included for the males. Increasing average annual hours of exposure to TSP above the various cutoffs were associated with increased risk of malignant neoplasms in males but not for females. Respiratory cancer risks were elevated but were not statistically significant. The increased risks for the males for the different cutoffs of TSP were of borderline statistical significance (p = .04-.06). There were no statistically significantly increased risks of malignant neoplasms for ozone in either males or females. Only one cutoff level approached statistical significance for the respiratory cancers (0,20; p = .07). Results of adjusting risk estimates for time spent indoors. The influence of time spent indoors on the air pollution-disease associations was accounted for by adjusting the information obtained on the 1977 questionnaire and by monthly adjustment factors provided by Winer et aI.l3 As expected, this resulted in a decrease in the range and mean values of both TSP and 03.For example, prior to adjustment for time spent indoors, the range of values for mean concentration of TSP was 0-157.21 pg/m3, and the mean value was 99.7 pg/m3; after adjustment, these values were 0-86.16 and 44.10 pg/m3, respectively. The values for mean concentration of O3 decreased from 0-4.21 pphm to 0-2.48 pphm after adjustment; the mean value decreased from 2.56 Archives of Environmental Health
Table 'I.-cox Proportional Hazards Regression for Respiratory Cancers, 1977-1982, with Hours Average Annual Concentration in Excess of 200 &m3 of Total Suspended Particulates (TSPs) as the Air Pollution Exposure Variable (n = 6 301, Cases = 17) Regression coefficient
Increment*
Relative riskt
95% C.I. for relative risk
TSP (hours in excess of 200 pg/rn')*
0.0005406
1 000hly
1.72
0.81, 3.65
Gender
0.9960
0.92, 7.98
0.03849 0.0246
(F, M) 10 Y
2.71
Total years smoked
Variable
Education
4Y
1.47
1.01, 2.14
1.10
0.60, 2.02
Downloaded by [University of Alberta] at 19:16 30 December 2014
'Increment for computations of relative risk. tRelative risk of increase in exposure of one increment, holding other variables in model constant. +Average annual hours in excess of 200 pg/m3, 1973-1977. gP < .05.
to 1.37 pphm, which indicated that most values decreased by approximately 50%. All mean concentration analyses were re-run after the indoor adjustments for TSP and O3 were obtained. In general, the coefficients adjusted for time spent indoors were larger than the unadjusted coefficients; e.g., for all malignant neoplasms among females, the coefficient associated with mean concentration of TSP increased from 0.0053 to 0.016 after adjustment for time spent indoors. For mean concentration of 0 3 associated with respiratory cancer, the coefficient increased from 0.4967 to 1.0048, which indicated that the unadjusted values gave conservative estimates of disease risk associated with long-term cumulative levels of TSP and 0,. Also, the result of the adjustment procedure was to improve the statistical significance of each outcome. Adjustment factors for the various cutoff levels could not be developed because there have been no studies of outdoor exposures in excess of cutoffs. Discussion
The associations between long-term ambient concentrations of TSP and O3 and cancer incidence and mortality rates in this study were observed in a unique population of California Seventh-day Adventists who were current nonsmokers. The lower cancer mortality rates of Adventi~ts'~may result in part from their lifestyles; they do not consume tobacco or alcohol, and they are vegetarians. Moreover, to ensure residential stability for a prolonged time period, participants in this study were required to have lived within 5 mi of their residence (in 1977) for at least 10 y. Reporting of cancer incidence in this population between 1977 and 1982 is nearly 100% c~mplete,'~and loss to follow-up has been minimal. In this population, increasing long-term ambient concentrations of TSP was associated with increased risk of all cancers combined among females, and the largest increases in risk estimates occurred among the smoking-related respiratory cancers (larynx, lung, pleura). Risk of all newly diagnosed cancers in females increased September/October 1991 [Vol. 46 (No. S)]
as cutoff levels of TSP increased, and it was statistically significant for all cutoffs of 100 &n3 or more, but this increase in risk was not observed among males. This excess risk among females was approximately twofold in the age-adjusted Mantel-Haenszel analysis and was of similar magnitude in the multivariate Cox analysis. The corresponding risk increases were threefold or greater for the smoking-related respiratory cancers. It is noteworthy that statistically significant elevations in cancer risk were observed at TSP100 but not at TSP75. This reinforces the prudence of maintaining the current standard for TSP at 75 &m3 because adverse health effects, at least in this population, must exist somewhere between TSP75 and TSP100, and it is best to err to the conservative side with regard to human health effects. There were statistically significant increases in respiratory risk for 03, and risk of all malignant neoplasms in females was elevated, although not significantly. Analysis of cancer mortality revealed somewhat stronger, though nonsignificant, results in the males than females. These results were observed for the mortality analysis after the follow-up period was increased by an additional 4 y. An explanation of the higher cancer incidence risk in females associated with TSP is not immediately apparent. Much smaller percentages of the females had been exposed to tobacco smoke and occupational fumes and dust. Only 14% of the females had ever smoked compared with 36% of the males. Moreover, only 5% of the females had smoked for at least 10 y, whereas 17% of the males had smoked at least that long. Also, only 1O/O of the females reported having ever worked in a hazardous occupation where exposure to fumes and dust was common (compared with 14% of the males). There is some evidence that the increased risk of lung cancer that is associated with urban living is more apparent in nonsmokers than in smoker^.'^'^^'' This is consistent with the females' stronger relationship between ambient concentrations of TSP and cancer incidence in this study. Moreover, females may have more exposure to ambient levels of TSP than do males because of their daily activity patterns. Whereas 277
Table 8.-Estimates of Cancer Incidence (1977-1982) Relative Risks from Cox Proportional Hazards Regressions for Incremental Increases of Exposure* above Various Cutoff levels of Total Suspended Particulates and for Mean Concentration of Total Suspended Particulates All malignant
Cutoff level
kdm7
Person-years in excess of increment
(Oh)
lncrementt size
neoplasms Males Females (n = 108) (n 175)
-
Downloaded by [University of Alberta] at 19:16 30 December 2014
~
60
99.7 88.1 77.50
75
98.6 78.8 67.5
100
83.0 43.0 24.6
150
200
Mean concentration
1 000 hly 3 500
Smoking-related respiratov cancers (n 17)
-
~
1.c4 1.13 1.19
1.05 1.18 1.27
1.21 1.96 2.62
5000
1.04 1.13 1.19
1.07 1.27 1.41
1.19 1 .a2 2.36
1 000 hly 2 500 5000
1.02 1.c4 1.08
1.10* 1.27$ 1.61
*
1.21 1.60 2.55
99.7 54.7 23.0
500 hly 1000 2 500
1.00
0.99
1.099 1.789 1.529
1.15 1.33 2.04
48.8 31.5 26.9 21.3
250 hly 500 750 1000
0.99 0.98 0.97 0.96
1.0811 1.1711 1.2711 1.3711
1.14 1.31 1.50 1.72
98.3 60.7
50 pdmJ 100
1.03 1.06
1.26 1.60
1.73 2.98
5000 1 000 hly
3 500
.m
1
‘Average annual concentration, January 1973 to March 1977. tlncrement for computations of relative risk. *p = .04. fjp = .01. Ilp < .02.
men may be more likely to go directly from their residence to a workplace (which, in California, is probably air conditioned), women are traditionally more likely to be engaged in various outdoor pursuits, e.g., errands, transporting children to and from school, etc. Particulate matter in ambient air contains substances that exhibit carcinogenic activity in experimental systems.’’ The polycyclic aromatic hydrocarbons have received the most attention; several are carcinogenic to animals and humans.lg Therefore, a direct relationship between increasing exposure to TSP and an increase in cancer incidence and mortality rates would be expected. A relationship between ambient sulfates and particulates and total mortality among 117 SMSAs in the United States in 1960 was suggested in earlier studies. One of the criteria proposed by Sir Austin B. Hill” for invoking a causal relationship between two variables is the presence of a biologic gradient. Such gradients were evident in our data. Even though the majority of studies that have attempted to evaluate the air pollution-cancer relationship have focused on lung cancer (e.g., reference number 22), the relationships with cancer at all sites and nonrespiratory tract cancers have been investigated in several other studies. For example, Winkelstein and K a n t 0 6 ~ found . ~ ~ that both stomach and prostate cancer mortality rates were higher in Buffalo, New York, where there was more TSP pollution than found in 278
other less-polluted areas. Other investigators noted significantly higher mortality rates for cancers of the stomach, esophagus, and bladder in more polluted areas of Nashville, Tennessee, than in less-polluted areas.25 In the present study, large increases in risk of respiratory cancer were associated with elevated TSP ambient concentrations, but increased risks for all malignant neoplasms were also observed, especially in females. These increased risks suggest that high ambient levels of TSP may have local and systemic effects on cancer induction. Tobacco smoke causes similar effects. It greatly enhances the risk of tumors that arise in the bronchial lining of the lung, which has direct, intimate contact with the smoke, but also enhances the risk of cancers in the pancreas and bladder, which are exposed to carcinogenic metabolites of various tobacco constituents.2“ To date, only one case-control study of air pollution and lung cancer has been rep~rted.~’ In that study of white males, cases (n = 417) and controls (n = 752) were selected from residents of areas with high, medium, or low TSP levels (maximum level was 200 pg/m3),. The author reports a nonsignificant odds ratio of 1.26 for residence in the high-pollution area. Despite these findings, the author noted that there was increased lung cancer risk from smoking and occupational exposure if there was also long-term exposure to air pollution. Our study has several strengths that were generally lacking in previous studies of relationships between Archives of Environmental Health
Table 9.-Cox Proportional Hazards Regression for Respiratory Cancer, 1977-1982, with Average Annual Hours in Excess of 10 pphm Ozone as the Air Pollution Exposure Variable (n = 6 301, Cases = 17) Regression coefficient
Increment*
Relative riskt
9 5 1 C.I. for relative risk
Oxidants (hours in excess of 10 pphm)* Gender
0.0016256§
500 hly
2.25
0.96, 5.31
(F, M)
3.62
1.16, 11.25
Total years smoked
0.0333
10 Y
1.40
0.93,
2.09
Education
0.0226
4Y
1.09
0.59,
2.03
Variable
~
i.2a5a11
~~~
for computations of relative risk. tRelative risk of increase in exposure of one increment, holding other variables in model constant. *Average annual hours in excess of 10 pphm, 1973-1977. gP < .05. llp ,055.
Downloaded by [University of Alberta] at 19:16 30 December 2014
-
concentrations of ambient air pollutants and cancer risks. The study was prospective, and it was based on a population of currently nonsmoking Seventh-day Adventists who were geographically stable for a period of at least 10 y prior to the beginning of follow-up for cancer incidence and mortality. These characteristics of the study population would have removed confounding that resulted from current smoking, and they would have limited any bias that resulted from migration patterns during the follow-up period (either outmigration to escape the relatively polluted areas under study or in-migration). It is possible, however, that individuals who were most sensitive to air pollution moved from higher pollution areas before the 10 y elapsed and were not included in our population. If this is true, the net result would be to bias our results toward the null. All incident cancers were diagnosed beginning in 1977, whereas ambient air pollution concentrations were ascertained beginning in 1966. The high response fraction of 87% made it unlikely that health characteristics that were associated with response were confounded with air pollution. Further assessment of this possibility for which response rates by geographic pollution area are compared is provided by Hodgkin.” We found that response fractions did not differ significantly between geographic areas of study. Many characteristics between high- and low-pollution geographic areas included in the study were compared by Hodgkin.*’ The majority of prior studies have relied upon inexact measures of ambient concentration levels for study subjects. There is an attempt to characterize individual experience in these studies by using census tract level measurements’ and county, and even statewide levels of measurements. The dangers of inferring exposuredisease relationships at the individual level by using aggregate exposure data are well known. We had the same problem to some degree. Analysis of respiratory cancer was plagued by the minimal number of events during the 6-y period of SeptembedOctober 1991 [Vol. 46 (No. 5)]
follow-up (n = 17). We plan to extend the mortality follow-up for an additional 4 y and to conduct a nested case-control study within the cohort with a larger number of respiratory cancer cases and an appropriate reference group. Another limitation of the present study is that ambient concentrations of TSP and 0, were evaluated but indoor concentrations were not. There are no known indoor sources of O,, and the indoor/outdoor concentration ratio for 0, has been estimated to range from 0.1 to 0.7.” It is likely that our population‘s actual exposures to 0, were considerably less than our estimates of ambient concentrations. Therefore, our findings of increased risk are most likely conservative with respect to magnitude of effect. We checked our results by restricting interpolations to be within the A or B quality ranges’ which are considered representative by the Environmental Protection Agency, and the final regression models were run using TSP200 and TSP mean concentration as the exposure variables. The numbers of individuals available for analysis were reduced by approximately one-half and by 6% for TSP and O, respectively. The regression coefficients were essentially unchanged for cancer outcomes after these restriction criteria were applied to both TSP and 0,. However, the adjustment factors are crude and in no way result in a representation of actual human exposures. It is not feasible to monitor human exposures in such a large-scale epidemiologic study during such a protracted period of time or to record the necessary information on personal habits each month for 20 y so that finer adjustments can be made. Indoor sources of TSP were modeled as covariates; the primary indoor source of TSP is tobacco smoke. In the final analysis, these adjustments did not change the health effects results, except to rescale the pollutant. The results of our study suggest that TSP may not be the best measure of toxic particulates suspected of causing adverse health effects. Currently, the fine279
Downloaded by [University of Alberta] at 19:16 30 December 2014
particulate fraction of TSP (< 2.5 p in aerodynamic diameter) is considered the most hazardous. Present plans for our ongoing prospective study call for the estimation of fine-particulate exposures and exposures to nitrogen dioxide, sulfur dioxide, and SO,. Finally, there are problems in separating statistically the effects of air pollutants, whose concentrations in the ambient air are highly correlated. Human chamber studies may untangle this multicollnearity, but such studies are not feasible for free-living populations. The process of relating health effects directly to community concentrations is relevant because air quality standards are based on ambient concentrations. In summary, we observed evidence that supports an association between increasing long-term cumulative levels of TSP and cancer incidence among females in our population. Also, there is suggestive, though not convincing, evidence that TSP is associated with respiratory cancer and that 0, may be related to respiratory cancer.
********** This research was supported by the California Air Resources Board, Grant numbers A6-128-33 and A833-057. Submitted for publication July 1, 1990; revised; accepted for publication April 16, 1991. Requests for reprints should be sent to: Paul K. Mills, Ph.D., Department of Health and Preventive Medicine, School of Medicine, Loma Linda University, Loma Linda, California 92350.
********** References
1. Stocks P, Campbell J. Lung cancer death rates among nonsmokers and pipe and cigarette smokers. An evaluation in relation to air pollution by benzpyrene and other substances. Br Med J 1955; 2~923-39. 2. Buffler P, Cooper S, Stinnett S, et al. Air pollution and lung cancer mortality in Harris County, Texas, 1979-81. Am J Epidemiol 1988; 128:683-99. 3. Euler GL, Abbey DE, Magie AR, Hodgkin JE. Chronic obstructive pulmonary disease symptoms effects of long-term cumulative exposure to ambient levels of total suspended particulates and sulfur dioxide in California Seventh-day Adventist residents. Arch Environ Health 1987; 42:213-22. 4. Detels R, Sayre JW, Coulson AH. The U.C.L.A. population studies of chronic obstructive respiratory disease. Am Rev Respir Dis 1981; 124:673-80. 5. Stocks P. Cancer and bronchitis mortality in relation to atmospheric deposit and smoke. Br Med J 1959; 74-79. 6. Zeidberg L. The Nashville air pollution study v. mortality from diseases of the respiratory system in relation to air pollution. Arch Environ Health 1967; 15:214-24. 7. U.S. Environmental Protection Agency. Air quality criteria for particulate matter and sulfur oxides. Publication no. EPA-600/882-029a. Research Triangle Park, NC: U.S. Environmental Protection Agency, 1982.
280
8. Abbey DE, Moore J, Petersen F, Beeson WL. Estimating cumulative ambient concentrations of air pollutants. Arch Environ Health 1991; 46:281-287. 9. Beeson WL, Fraser GE, Mills PK. Validation of record linkage to two California population-based tumor registries in a cohort study. Proceedings of the 1989 Public Health Conference on Records and Statistics. DHHS Publication no. (PHS)90-1214, 1990; pp. 196-201. 10. Cox DR. Regression models and life tables. J Royal Stat SOC 1972; 34 Ser B:187-220. 11. Dixon WJ, Chief Ed. BMDP statistical software, 1983 printing with additions. Berkeley: University of California Press. 12. Breslow NE, Day NE, Statistical methods in cancer research. Vol II. The design and analysis of cohort studies. IARC Scientific Publications no. 82. Lyon: International Agency for Research on Cancer, 1987; p. 180. 13. Winer AM, Lurmann FW, Coyner LA, Colorne SD, Poe MP. Characterization of air pollution exposures in the California South Coast Air Basin: application of a new regional human exposure (rehex) model. In: Final report, SCAQMD, Statewide Air Pollution Research Center, 1989; pp. 33-47. 14. Phillips RL, Garfinkel L, Kuzma J et al. Mortality among California Seventh-day Adventists for selected cancer sites. J Natl Cancer lnst 1980 65:1097-1107. 15. Beeson WL, Mills PK, Phillips RL, Andress M, Fraser GE. Chronic disease among Seventh-day Adventists, a low risk group. Rationale, methodology and description of the population. Cancer 1989; 64:570-81. 16. Dean G . Lung cancer and bronchitis in Northern Ireland. Br Med J 1966; 1:1506-14. 17. Buell P, Dunn JE, Breslow L. Cancer of the lung and Los Angelestype air pollution. Cancer 1967; 20:2139-47. 18. Hoffman D, Wynder EL. Organic particulate pollutants. In Stern AC, Ed. Air pollution, vol 11, 3d ed. New York: Academic Press, 1977; pp. 361-55. 19. National Research Council. Particulate polycyclic organic matter. Washington, DC: Committee on Medical and Biological Effects of Environmental Pollutants, National Academy of Sciences, 1972. 20. Lave LB, Seskin EP. Air pollution and human health. Baltimore, MD: The Johns Hopkins University Press, 1977; p. 44. 21. Hill AB. The environment and diseases: association and causation. Proc Royal SOC Med (section on Occupational Medicine) 1965; 58:272. 22. Mills CA. Urban air pollution and respiratory diseases. Am J Hygiene 1943; 37:131. 23. Winkelstein W Jr, Kantor S. Prostatic cancer: relationship to suspended particulate air pollution. Arch Environ Health 1969; 18:544. 24. Winkelstein W Jr, Kantor S. Stomach cancer: positive association with suspended particulate air pollution. Arch Environ Health 1969; 18:544. 25. Hagstrum RM, Sprague HA, Landau E. The Nashville air pollution study. VII. Mortality from cancer in relation to air pollution. Arch Environ Health 1967; 15:237. 26. Wynder E, Hoffman D. Tobacco. In Schottenfield D, Fraumeni JF Jr, Eds. Cancer epidemiology and prevention. Philadelphia, PA: W. B. Saunders, 1982. 27. Vena J. Air pollution as a risk factor for lung cancer. Am I Epidemiol 1982; 1 16:42-56. 28. Hodgkin JE, Abbey DE, Euler CL, Magie AR. COPD prevalence in nonsmokers in high and low photochemical pollution areas. Chest 1986; 86(6):830-38. 29. Colome SD, Wilson AL, Becker EW, Cunningham SJ, Baker PE. Analysis of factors associated with indoor residential nitrogen dioxide: multivariate analysis. Indoor Air 1987. Proceedings of the 4th International conference on indoor Air Quality and Climate 1987; 1:405-09.
Archives of Environmental Health
Archives of Environmental Health: An International Journal Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/vzeh20
Ambient Air Pollution and Cancer in California Seventh-day Adventists a
a
a
Paul K. Mills Ph.D., M.P.H. , David Abbey Ph.D. , W. Lawrence Beeson M.S.P.H. & Floyd Petersen M.P.H.
a
a
Department of Public Health and Preventive Medicine , School of Medicine, Loma Linda University , Loma Linda, California, USA Published online: 03 Aug 2010.
To cite this article: Paul K. Mills Ph.D., M.P.H. , David Abbey Ph.D. , W. Lawrence Beeson M.S.P.H. & Floyd Petersen M.P.H. (1991) Ambient Air Pollution and Cancer in California Seventh-day Adventists, Archives of Environmental Health: An International Journal, 46:5, 271-280, DOI: 10.1080/00039896.1991.9934387 To link to this article: http://dx.doi.org/10.1080/00039896.1991.9934387
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
Ambient Air Pollution and Cancer in California Seventh-day Adventists
Downloaded by [University of Alberta] at 19:16 30 December 2014
PAUL
K. MILLS,
Ph.D.,
M.P.H.
DAVID ABBEY, Ph.D. W. LAWRENCE BEESON, M.S.P.H. FLOYD PETERSEN, M.P.H. Department of Public Health and Preventive Medicine School of Medicine Loma Linda University Lorna Linda, California
ABSTRACT. Cancer incidence and mortality in a cohort of 6 O00 Seventh-day Adventist nonsmokers who were residents of California were monitored for a b y period, and relationships with long-term ambient concentrations of total suspended particulates (TSPs) and ozone (0,)were studied. Ambient concentrations were expressed as mean concentrations and exceedance frequencies, which are the number of hours during which concentrations exceeded specified cutoffs (e.g., federal and California air quality standards). Risk of malignant neoplasms in females increased concurrently with exceedance frequencies for all TSP cutoffs, except the lowest, and these increased risks were highly statistically significant. An increased risk of respiratory cancers was associated with only one cutoff of 4, and this result was of borderline significance. These results are presented in the context of setting standards for these two air pollutants.
cross-sectional or correlational d e ~ i g n . ~In, ~ this apADVERSE HEALTH EFFECTS associated with ambient proach, indices of air pollution in certain geographic air pollution have been scrutinized by environmental locales were correlated with the age-adjusted cancer scientists for most of the twentieth century.',2 There is mortality rates in the same areas, with or without adevidence that air pollution contributes to morbidity from airway obstructive disease and other forms of resjustment for consumption of cigarettes. Most of these piratory disease, cancer, and cardiovascular di~ease.~.~ studies have not demonstrated that an increased cancer risk occurs with an increase in air pollution levels. It is difficult to evaluate this evidence because the efIn fact, the U.S. Environmental Protection Agency crifects of tobacco smoke are difficult to separate from the effects of air pollution, especially in urban areas where teria document for particulate matter and sulfur oxides the numbers of smokers and concentrations of air polconcluded: ". . . nor does there presently exist credible lutants are higher than in more rural areas. Cigarette epidemiological evidence linking increased cancer smoking has been implicated in the etiology of cardiorates to elevations in particulate matter as a class, i.e., undifferentiated as to chemical content."' In many invascular disease, respiratory diseases, lung cancer, and several other forms of cancer in humans (e.g., cancers stances, however, an inability to detect relationships can be linked with limitations in study design, poor of the bladder and pancreas). Most studies of air pollution and cancer in humans statistical power, or other methodological weaknesses. Seventh-day Adventists provide a unique opportunity have been epidemiologic investigations that used SeptemberlWober1991 [Vol. 46 (NO.5)]
271
for investigating the health effects of ambient air pollutants with very little confounding by tobacco smoke. Use of tobacco or alcohol is proscribed by the Seventh-day Adventist church. However, Seventh-day Adventists‘ contact with ambient air pollution varies greatly by virtue of their choice of residence and occupation. This variability in pollution exposure makes possible the evaluation of the relationship between exposure and health effects, and there is minimal confounding by tobacco smoke. In this report, we present the risks of cancer associated with long-term exposure to ambient concentrations of total suspended particulates (TSP) and ozone (03), which were observed in a cohort study of California Seventh-day Adventists who did not smoke.
Downloaded by [University of Alberta] at 19:16 30 December 2014
Methods The study design and measurement of ambient air pollution have been described in a previous study.’ Enrollment of the study population. In April of 1977, a subgroup of 6 340 members of the National Cancer Institute-funded Adventist Health Study were enrolled in a prospective epidemiological study for the primary purpose of studying the health effects of long-term cumulative contact with air pollution. Persons who reported that they were current smokers were excluded from the study. Because many individuals had not been Adventists all their lives, they may have smoked in the past or may have resided with a non-Adventist who smoked. Therefore, detailed smoking histories were ascertained for each participant. Individuals who met the following criteria were included in the study: (1) age 25 y+ at the time the Adventist Health Study questionnaire was completed in 1974; (2) member of the Seventh-day Adventist church at the time of enrollment in the Adventist Health Study; (3) Non-Hispanic white (individuals of all races completed the questionnaire, but only non-Hispanic whites were followed for cancer and heart disease surveillance); (4) lived at least 10 y within 5 mi of their current residence; and (5) resided in one of three metropolitan areas (San Francisco, Los Angeles and eastward [South Coast Air Basin], or San Diego), or were included in a random sample of 862 individuals who resided in the remainder of California. These 862 individuals represented a 10% random sample of Adventist Health Study participants who met all of the above criteria, except for number 5. Individuals who met these criteria received, by mail, a baseline questionnaire that ascertained their residence history by month and zip code since 1960 and their lifestyle habits pertinent to relative air pollution exposure, e g , work location, hours driving on crowded freeways, percentage of time spent indoorsloutdoors, etc. They also completed the National Heart, Lung, and Blood Institute’s (NHLBl’s) respiratory symptoms questionnaire, which was used to determine selfreported symptoms of chronic respiratory disease. Detailed smoking histories were also obtained as were histories of ever having lived or worked with a smoker and the duration of those exposures. The response fraction to the baseline questionnaire was 87%. A wide range of 272
other lifestyle data (e.g., demographic and dietary information) were already available for these individuals because they were included in the Adventist Health Study database. Cancer monitoring program
Incidence. A surveillance system consisted of annual mailings to every member of the cohort between April 1, 1977, and December 31, 1982. Information on any hospitalization that may have occurred during the prior 12-mo period of follow-up was obtained. If a hospitalization was reported, the name and address of the respective hospital were recorded, and permission to review the resulting medical record was obtained. Adventist Health Study personnel reviewed all medical records for a diagnosis of cancer or cardiovascular disease. Pertinent portions of the medical records were microfilmed to allow a confirmation of the diagnosis by senior medical personnel. This type of follow-up was complete for 99% of the cohort. Computerized record linkage was conducted in areas of California where there were population-based tumor registries (i.e., Cancer Surveillance Program, Los Angeles County; the Resource for Cancer Epidemiology, San Francisco Bay area).g Mortality. Three mechanisms were used to monitor mortality in the study population: (1) computerized record linkage with the California death certificate file, (2) computerized record linkage with the National Death Index, and (3) manual linkage with Seventh-day Adventist church records. These mechanisms enabled us to identify deaths that occurred in the study population during the period of follow-up (1977-1986). Statistical methods Adjusted relative risks (ratios of incidence rates) are presented for average annual exposures to TSP and O3 that exceeded cutpoints versus lower exposures. Relative risks were adjusted for relevant covariates, including age, sex, and past smoking history. Mantel-Haenszel analyses were performed for each outcome before the more sophisticated analyses with proportional hazards regression models. These analyses categorized continuous variables, and any assumptions of linear or additive effects were avoided. The disadvantages of these analyses are the loss of statistical power resulting from this categorization and their inability to adjust for more than a few covariates (two covariates, with the sample size of the present study). The results of the Mantel-Haenszel analysis should not be considered conclusive, but should be used only as a check on the multivariate models. The Mantel-Haenszel analyses were age-adjusted according to the following strata: 25-44 y, 45-64 y, 65-79 y, and 80 y and older. Cumulative annual average concentrations and their annual average exceedance frequencies above several cutoffs were used as indices for long-term ambient concentrations of TSP and 03.Two cutpoints for each index of exposure were determined from the cumulative distributions of the indexes for the Archives of EnvironmentalHealth
Downloaded by [University of Alberta] at 19:16 30 December 2014
subcohorts who resided inside and outside the South Coast Air Basin of California (SCAB). Not much overlap existed between the distributions of the generally lower ambient concentrations outside the SCAB and ambient concentrations inside the SCAB. The lower of the two cutpoints was the 90th percentile of ambient concentrations outside the SCAB, and the higher cutpoint was the median of all inside SCAB concentrations that were above the lower cutpoint. Cox proportional hazards regression models for cancer incidence and mortality outcomes, which controlled for several covariates, were fitted to the data. Cox proportional hazards regression models” are the model of choice for cancer incidence and mortality outcomes when time to event in the cohort is the variable of primary interest. These models controlled for several covariates and were fitted by the BMDP2L stepwise procedure.” Cumulative ambient concentrations of TSP and 0, were represented by annual average concentrations and by annual average exceedance frequencies that were above cutoffs for two time periods (January 1966 to March 1977 and January 1973 to March 1977). The second time period was allowed as an alternative for the longer time periods in the models. We thought that ambient pollutant monitoring was more representative of that experienced in the locations of study participants during this later period because of an increased number of monitoring stations. Because all study participants had lived in their neighborhoods for at least 10 y prior to 1977, we thought the later time period might provide a better ranking of the relative ambient concentrations that were experienced by study subjects for the longer time period. Cumulative ambient concentrations for the two time periods were allowed to compete for entry in the stepwise selection process that was used for original model formulation. Separate models were fitted for annual average concentrations and for exceedance frequencies that were above each cutoff. Time-dependent Cox regression analyses were also performed; average pollutant exposures between January 1973 and the time of risk set as the exposure variables were used. Models that involved ambient concentrations of TSP and 0, were fitted separately. Models were first fit for TSP using exceedance frequencies for the cutoff of 200 pg/m3 (TSP200) and ozone using exceedance frequencies for the cutoff of 10 pphm (0,lO). For each pollutant the variables selected for these cutoffs were then used for the cutoffs as well as for mean concentration. The primary candidate independent variables-total years of smoking, gender, and education-were forced into each model. Age was not among the covariates because age, instead of time in study, was used as the time variable in the models (as recommended by Breslow”). The use of age as the time variable provided a tighter control and adjustment for age. This was especially important for disease outcomes that were strongly age-related. The secondary candidate independent variables that were considered for entry into the models were years lived with a smoker, years worked with a smoker, and past or present employment in an occupation during which high exposures to airborne contaminants occurred. A list of the occupa!kptember/October1991 [Vol. 46 (No.S)]
tions in which the study participants experienced high exposures is given in the study by Euler et aL3 The pollutant exposure variable or variables and the secondary candidate variables competed for entry by the stepwise selection procedure. An F to remove of 0.15 and an F to enter of 0.10 were used to terminate the stepwise procedure. The final models chosen required an F to enter of 0.05 or less. Interactions between TSP or O3 and the other covariates were assessed; first-order cross-product terms in the model were included and then removed, and the change in the log-likelihood was noted.
Results Three different disease outcomes were analyzed in this analysis: (1) all malignant neoplasms for males, (2) all malignant neoplasms for females, and (3) respiratory cancers in all subjects. All malignant neoplasms in males and females were analyzed separately because some cancer sites are sex specific. The distributions of the relevant covariates (e.g., past history of smoking, education) are shown in Table 1. Approximately 62% of the study population resided within the South Coast Air Basin. The frequency distributions of the air pollution variables are shown in Table 2. Between the initiation of follow-up on April 1, 1977, and the end of incidence follow-up on Decernber 31, 1982, there were 301 newly diagnosed cancers (ICDO 140-200) detected in the population (Table 3). Only 17 respiratory cancers were diagnosed. Several demographic and smoking characteristics of these respiratory cancer cases are presented in Table 4. During the mortality follow-up period (1977-19861, there were 180 cancer deaths.
Cancer incidence Results of the stepwise regressions indicated that exceedance frequencies during January 1973 through March 1977 were most closely associated with cancer risk between 1977 and 1982. For this reason, average annual exceedance frequencies during this time period were use in the regression analyses. However, exceedance frequencies for the longer time period (1966-1977) were used to check all results, and only minor differences in the results were noted. The Cox proportional hazards regression models, which were fitted, each included exceedance frequency variables for a cutoff of TSP or 0, for covariates that included gender, education, total years of past smoking, and past or present ernployment in occupations that involved exposure to airborne contaminants. (Occupational exposures were included only for males because females lacked sufficient exposure.) Passive smoking variables did not contribute sufficiently to the fit of the model to merit inclusion. The result of the multivariate analysis of TSP2OO is shown in Table 5. For 1 OOO h/y in excess of this cutoff, the relative risk for all malignant neoplasms among females was 1.37 and was statistically significant fp < .05). Although the relative risk for malignant neoplasms in females associated with mean concentration of TSP was elevated, this was not statistically significant. An 273
examination of the organ site-specific cancer sites that contributed to this excess is presented in Table 6 and indicates that risk of breast cancer was elevated (relative risk [RR] = 1.511, as was risk in all other cancer sites combined (RR = 1.65). Neither of these risk estimates was statistically significant. The relative risk for respiratory cancers for 1 O00 h in excess of TSP200 was 1.72 (Table 7); however, this risk is not statistically significant. Again, the risk associated with mean concen-
Table 1.-Frequency Distributions of Past and Passive Smoking Variables and Covariates in the AHSMOG Incidence Cohort
Downloaded by [University of Alberta] at 19:16 30 December 2014
Variable History of smoking Never Past only
Females
Males
(n = 4 063) NO. %
(n = 2 277) NO.
%
3 504 559
86 14
1 450
827
64 36
3 475 335 196 57
86 08 05 01
1484 339 390 64
675 15 17 03
3 519 292 246 6
87 07
1 507 312 429 30
66 14 19 01
698 1 863 1 149 353
17 46 28 09
358 1170 599 150
16 51 26 07
Education Some high school High school graduate Some college College graduate Unknown
868 695 1715 760 25
21 17 42 19 01
41 7 249 743 856 12
18 11 33 38 01
Occupational air pollution No Yes
4 026 37
99 01
1 964
86 14
2 119 403 1 541 0
52 10 38 0
1 508 236 533 0
66 10 23 0
2 512 815 736
62 20 18
1183 392 702
52 17 31
0
0
0
0
Pack years of cigarette smoking None
< 10
B 10 Unknown Total years of smoking Never 1-9 y
B 1oy Unknown Age (as of 4/1/77) 25-44 y 45-64 y 65-80 y
Years lived with a smoker Never 1-9 y
2 l0y Unknown Years worked with a smoker Never 1-9 y
B l0y Unknown Residence in South Coast Air Basin* Outside Inside Mixed
06
01
Days
Hours" x 1000
313
RR =1.4'
\
TSP THRESHOLD (ug/m3)
* 2 403 3 914 384
32 62 06
*Residence inside or outside the South Coast Air Basin, 1973-1977. Numbers include males and females.
274
tration of TSP was elevated but was not statistically significant. Relative risk contour plots for the associations between all malignant neoplasms in females and TSP exposures are shown in Figure 1. The relative risks for all cutoffs, except for TSP60 and TSP75, are statistically significant. The "dose-response"-type curves for the various cutoffs of TSP are displayed in Figure 2. The results of the multivariate TSP analyses for the other cutoffs and for mean concentration are presented in Table 8. Separate Cox regressions were used for each cutoff and were used with the same covariates as for TSP200. Statistically significant increases in cancer risk were seen among females for all cutoffs, except for TSP6O and TSP75. The magnitude of the increased risk for females is approximately 40-60%. Twofold increases in risk were observed for respiratory cancers, but these risks were not statistically significant. Again there was no relationship observed between TSP ambient concentrations and cancer incidence in the males. The results of the Cox regressions that relate respiratory cancer incidence and O3 ambient concentrations are presented in Table 9. For the 500 h in excess of 0310, the relative risk for respiratory cancer was 2.25, which is of borderline statistical significance. Statistical significance was not achieved for any of the other cutoffs or for mean concentration. The results of the Cox proportional hazards regression analysis were very sim-
**
Risks for TSP60 and TSWS not statistically significant. Average annual hours 1973 - 1977.
Fig. 1. Relative risk contour plots for all malignant neoplasms (female) associated with TSP. Contour plots show incremental increases in hours (days) in excess of different cutoffs required to p m duce stated relative risk.
Archives of Environmental Health
Table 2.-Frequency
Distribution of Total Suspended Particulates and Ozone in the AHSMOG Incidence Cohort
Downloaded by [University of Alberta] at 19:16 30 December 2014
Average annual exposure*
No.
YO
Average annual exposure
TSP60 (hly) Unknown (6000 6 OO1-7 500 >7500
204 1 986 2 429 1721
03 31 38 27
TSP75 (hly) Unknown 6 725
204 2 078 1917 1 743
TSPl00 (hly) Unknown (3000 3 001-5 000 >5000 TSP150 (hly) Unknown < 750 751-2 000 >2000 TSP200 (hly) Unknown < 125 126-1 000 >lo00 TSP mean concentration hg/m3) < 92.41 92.41-114.37 > 114.37
No.
010
0310 (hly) Unknown 600
37 2 216 2 291 1 776
01 35 36 28
03 35 32 29
0312 (hly) Unknown < 25 25-400 >400
37 2 132 2 209 3 745
01 26 27 46
204 2 259 2 288 1 589
03 36 36 25
0315 (hly) Unknown 200
37 2 260 2 305 1 735
01 36 36 27
204 2 237 2 184 1715
03 35 35 27
0320 (h/y) Unknown 50
37 2 260 2 074 1 966
01 35 33 31
204 1 996 2 801 1 339
03 32
44 21
37 2 260 1 585 2 455
01 36 25 38
2 119 1956 2 028
35 32 33
2 779 1 824 1455
46 30 24
0325 (Illy) Unknown 0 %lo > 10 0, mean concentration (pphm) < 2.5634 2.5634-3.1000 > 3.1000
*Average annual hours of exposure to indicated thresholds of total suspended particulates (e.g., 60 or more pg/m3) 1973 to earliest of date of censoring, date of death, or December 31, 1982.
Table 4.-Demographic and Smoking-Related Attributes of Respiratory Cancer Cases in the AHSMOG Incidence Population, 1977-1982
Table J.-Sex-Specific Distribution of Organ Sitespecific Cancer in the AHSMOG Incidence Population Females
Males
Site Colon Rectum Pancreas Lung Leukemia Skin Breast Cervix Uterine Ovary Thyroid Bladder Lymphoma Prostate Total
22 12 8 5 6 9 65 5 37 10 4 0 6 0 189
12 6 4 3 3 5 34 3 20 5 2 0 3 0 100
17 3 0 12 6 16 0 0 0 0 0 7 9 42 112
15 3 0 11 5 14 0 0 0 0 0 6 8 38 100
ilar to the results obtained in the Mantel-Haenszel stratified analysis. Only age and sex were controlled in the stratified analysis. Multipollutant analysis. The relative contributions of September/October 1991 [Vol. 46 (No. 5)]
Attribute
n
Sex Male Female
12 5
Age (y)
(64 65-74 75-84 > 85 Education (8 9-1 2 < 16 Cancer site Female Bronchus Lung Male Lung Larynx Pleura
4 4 8 1 3 5 3
1
Attribute
n
Total years smoked Never 10 < 10 2 > 10 5 Years lived with smokers Never 9 < 10 1 > 10 7 Years worked with smokers Never 10 6 10 1 > 10 6 Hazardous occupation No 15 Yes 2
4
8 2 2
275
Table 5.-cox Proportional Hazards Regression for All Malignant Neoplasms among Females, 1977-1982, with Hours Average Annual Concentration in Excess of ZOO pg/m3 of Total Suspended Particulates USPs) as the Air Pollution Exposure Variable (n = 4 063, Cases = 175)
I
Variable
Regression coefficient
Increment *
Relative riskt
95% C.I. for relative risk
TSP (hours in excess
0.00031745
1 000h/y
1.37
1.05, 1.80
of 200 pg/mW Total years smoked
0.0223
1OY
1.25
0.98, 1.59
Education
0.0424
4Y
1.18
0.94, 1.50
Downloaded by [University of Alberta] at 19:16 30 December 2014
'Increment for computations of relative risk. tRelative risk of increase in exposure of one increment, holding other variables in model constant. *Average annual hours in excess of 200 pg/m3, 1973-1977. §p < .05.
~~
~
~~
Table 6.-Cox Proportional Hazards Regression for Site Specific Cancers among Females Associated with TSPZOO in Excess of 1 OOO h/y (1973-1977) in the AHSMOG Incidence Population ~~
Organ site Breast Cervix Colorectal Respiratoy Leukemiallymphoma All other cancers
Relative risk (95% C.I.)
p Value
1.51 (0.92-2.47) 0.64 (0.05-7.55) 1.08 (0.55-2.1 1) 0.72 (0.12-4.50) 1.05 (0.33-3.37) 1.65 (0.90-3.01)
.10 .72 .83
.72 .94 .10
1.8 -
5000 hrslyr
&
1.6 1.4
1.2
1
TSP THRESHOLD (ug/m3)
*
Risks for TSP60 and TSP75 not statistically significant.
Fig. 2. Relative risk of all malignant neoplasms among females for differing annual average hours (1973-1977) in excess of 60, 75, 100, 150, and 200 pg/m3 of total suspended particulates.
276
TSP and 0, to cancer risk were evaluated. We compared the contributions of the TSP and O3 cutoffs most strongly associated with risk in both the MantelHaenszel and the Cox regression analyses, i.e., 200 pg/m3 and 10 pphm, respectively. With respect to respiratory cancer, 0310 entered the model with a p-toenter value, which was of borderline statistical significance (p = .055), and TSP200 did not enter. Conversely, for all malignant neoplasms among females, TSP 200 entered the model (p to enter = .038), but 0 , l O did not. Cancer mortality. The same Cox proportional hazards regression models were fit to the mortality data as were fit to the incidence data. Covariates included total years of smoking and education (as a surrogate measure of social class). Past or present experience in an occupation with high levels of airborne contaminants was also included for the males. Increasing average annual hours of exposure to TSP above the various cutoffs were associated with increased risk of malignant neoplasms in males but not for females. Respiratory cancer risks were elevated but were not statistically significant. The increased risks for the males for the different cutoffs of TSP were of borderline statistical significance (p = .04-.06). There were no statistically significantly increased risks of malignant neoplasms for ozone in either males or females. Only one cutoff level approached statistical significance for the respiratory cancers (0,20; p = .07). Results of adjusting risk estimates for time spent indoors. The influence of time spent indoors on the air pollution-disease associations was accounted for by adjusting the information obtained on the 1977 questionnaire and by monthly adjustment factors provided by Winer et aI.l3 As expected, this resulted in a decrease in the range and mean values of both TSP and 03.For example, prior to adjustment for time spent indoors, the range of values for mean concentration of TSP was 0-157.21 pg/m3, and the mean value was 99.7 pg/m3; after adjustment, these values were 0-86.16 and 44.10 pg/m3, respectively. The values for mean concentration of O3 decreased from 0-4.21 pphm to 0-2.48 pphm after adjustment; the mean value decreased from 2.56 Archives of Environmental Health
Table 'I.-cox Proportional Hazards Regression for Respiratory Cancers, 1977-1982, with Hours Average Annual Concentration in Excess of 200 &m3 of Total Suspended Particulates (TSPs) as the Air Pollution Exposure Variable (n = 6 301, Cases = 17) Regression coefficient
Increment*
Relative riskt
95% C.I. for relative risk
TSP (hours in excess of 200 pg/rn')*
0.0005406
1 000hly
1.72
0.81, 3.65
Gender
0.9960
0.92, 7.98
0.03849 0.0246
(F, M) 10 Y
2.71
Total years smoked
Variable
Education
4Y
1.47
1.01, 2.14
1.10
0.60, 2.02
Downloaded by [University of Alberta] at 19:16 30 December 2014
'Increment for computations of relative risk. tRelative risk of increase in exposure of one increment, holding other variables in model constant. +Average annual hours in excess of 200 pg/m3, 1973-1977. gP < .05.
to 1.37 pphm, which indicated that most values decreased by approximately 50%. All mean concentration analyses were re-run after the indoor adjustments for TSP and O3 were obtained. In general, the coefficients adjusted for time spent indoors were larger than the unadjusted coefficients; e.g., for all malignant neoplasms among females, the coefficient associated with mean concentration of TSP increased from 0.0053 to 0.016 after adjustment for time spent indoors. For mean concentration of 0 3 associated with respiratory cancer, the coefficient increased from 0.4967 to 1.0048, which indicated that the unadjusted values gave conservative estimates of disease risk associated with long-term cumulative levels of TSP and 0,. Also, the result of the adjustment procedure was to improve the statistical significance of each outcome. Adjustment factors for the various cutoff levels could not be developed because there have been no studies of outdoor exposures in excess of cutoffs. Discussion
The associations between long-term ambient concentrations of TSP and O3 and cancer incidence and mortality rates in this study were observed in a unique population of California Seventh-day Adventists who were current nonsmokers. The lower cancer mortality rates of Adventi~ts'~may result in part from their lifestyles; they do not consume tobacco or alcohol, and they are vegetarians. Moreover, to ensure residential stability for a prolonged time period, participants in this study were required to have lived within 5 mi of their residence (in 1977) for at least 10 y. Reporting of cancer incidence in this population between 1977 and 1982 is nearly 100% c~mplete,'~and loss to follow-up has been minimal. In this population, increasing long-term ambient concentrations of TSP was associated with increased risk of all cancers combined among females, and the largest increases in risk estimates occurred among the smoking-related respiratory cancers (larynx, lung, pleura). Risk of all newly diagnosed cancers in females increased September/October 1991 [Vol. 46 (No. S)]
as cutoff levels of TSP increased, and it was statistically significant for all cutoffs of 100 &n3 or more, but this increase in risk was not observed among males. This excess risk among females was approximately twofold in the age-adjusted Mantel-Haenszel analysis and was of similar magnitude in the multivariate Cox analysis. The corresponding risk increases were threefold or greater for the smoking-related respiratory cancers. It is noteworthy that statistically significant elevations in cancer risk were observed at TSP100 but not at TSP75. This reinforces the prudence of maintaining the current standard for TSP at 75 &m3 because adverse health effects, at least in this population, must exist somewhere between TSP75 and TSP100, and it is best to err to the conservative side with regard to human health effects. There were statistically significant increases in respiratory risk for 03, and risk of all malignant neoplasms in females was elevated, although not significantly. Analysis of cancer mortality revealed somewhat stronger, though nonsignificant, results in the males than females. These results were observed for the mortality analysis after the follow-up period was increased by an additional 4 y. An explanation of the higher cancer incidence risk in females associated with TSP is not immediately apparent. Much smaller percentages of the females had been exposed to tobacco smoke and occupational fumes and dust. Only 14% of the females had ever smoked compared with 36% of the males. Moreover, only 5% of the females had smoked for at least 10 y, whereas 17% of the males had smoked at least that long. Also, only 1O/O of the females reported having ever worked in a hazardous occupation where exposure to fumes and dust was common (compared with 14% of the males). There is some evidence that the increased risk of lung cancer that is associated with urban living is more apparent in nonsmokers than in smoker^.'^'^^'' This is consistent with the females' stronger relationship between ambient concentrations of TSP and cancer incidence in this study. Moreover, females may have more exposure to ambient levels of TSP than do males because of their daily activity patterns. Whereas 277
Table 8.-Estimates of Cancer Incidence (1977-1982) Relative Risks from Cox Proportional Hazards Regressions for Incremental Increases of Exposure* above Various Cutoff levels of Total Suspended Particulates and for Mean Concentration of Total Suspended Particulates All malignant
Cutoff level
kdm7
Person-years in excess of increment
(Oh)
lncrementt size
neoplasms Males Females (n = 108) (n 175)
-
Downloaded by [University of Alberta] at 19:16 30 December 2014
~
60
99.7 88.1 77.50
75
98.6 78.8 67.5
100
83.0 43.0 24.6
150
200
Mean concentration
1 000 hly 3 500
Smoking-related respiratov cancers (n 17)
-
~
1.c4 1.13 1.19
1.05 1.18 1.27
1.21 1.96 2.62
5000
1.04 1.13 1.19
1.07 1.27 1.41
1.19 1 .a2 2.36
1 000 hly 2 500 5000
1.02 1.c4 1.08
1.10* 1.27$ 1.61
*
1.21 1.60 2.55
99.7 54.7 23.0
500 hly 1000 2 500
1.00
0.99
1.099 1.789 1.529
1.15 1.33 2.04
48.8 31.5 26.9 21.3
250 hly 500 750 1000
0.99 0.98 0.97 0.96
1.0811 1.1711 1.2711 1.3711
1.14 1.31 1.50 1.72
98.3 60.7
50 pdmJ 100
1.03 1.06
1.26 1.60
1.73 2.98
5000 1 000 hly
3 500
.m
1
‘Average annual concentration, January 1973 to March 1977. tlncrement for computations of relative risk. *p = .04. fjp = .01. Ilp < .02.
men may be more likely to go directly from their residence to a workplace (which, in California, is probably air conditioned), women are traditionally more likely to be engaged in various outdoor pursuits, e.g., errands, transporting children to and from school, etc. Particulate matter in ambient air contains substances that exhibit carcinogenic activity in experimental systems.’’ The polycyclic aromatic hydrocarbons have received the most attention; several are carcinogenic to animals and humans.lg Therefore, a direct relationship between increasing exposure to TSP and an increase in cancer incidence and mortality rates would be expected. A relationship between ambient sulfates and particulates and total mortality among 117 SMSAs in the United States in 1960 was suggested in earlier studies. One of the criteria proposed by Sir Austin B. Hill” for invoking a causal relationship between two variables is the presence of a biologic gradient. Such gradients were evident in our data. Even though the majority of studies that have attempted to evaluate the air pollution-cancer relationship have focused on lung cancer (e.g., reference number 22), the relationships with cancer at all sites and nonrespiratory tract cancers have been investigated in several other studies. For example, Winkelstein and K a n t 0 6 ~ found . ~ ~ that both stomach and prostate cancer mortality rates were higher in Buffalo, New York, where there was more TSP pollution than found in 278
other less-polluted areas. Other investigators noted significantly higher mortality rates for cancers of the stomach, esophagus, and bladder in more polluted areas of Nashville, Tennessee, than in less-polluted areas.25 In the present study, large increases in risk of respiratory cancer were associated with elevated TSP ambient concentrations, but increased risks for all malignant neoplasms were also observed, especially in females. These increased risks suggest that high ambient levels of TSP may have local and systemic effects on cancer induction. Tobacco smoke causes similar effects. It greatly enhances the risk of tumors that arise in the bronchial lining of the lung, which has direct, intimate contact with the smoke, but also enhances the risk of cancers in the pancreas and bladder, which are exposed to carcinogenic metabolites of various tobacco constituents.2“ To date, only one case-control study of air pollution and lung cancer has been rep~rted.~’ In that study of white males, cases (n = 417) and controls (n = 752) were selected from residents of areas with high, medium, or low TSP levels (maximum level was 200 pg/m3),. The author reports a nonsignificant odds ratio of 1.26 for residence in the high-pollution area. Despite these findings, the author noted that there was increased lung cancer risk from smoking and occupational exposure if there was also long-term exposure to air pollution. Our study has several strengths that were generally lacking in previous studies of relationships between Archives of Environmental Health
Table 9.-Cox Proportional Hazards Regression for Respiratory Cancer, 1977-1982, with Average Annual Hours in Excess of 10 pphm Ozone as the Air Pollution Exposure Variable (n = 6 301, Cases = 17) Regression coefficient
Increment*
Relative riskt
9 5 1 C.I. for relative risk
Oxidants (hours in excess of 10 pphm)* Gender
0.0016256§
500 hly
2.25
0.96, 5.31
(F, M)
3.62
1.16, 11.25
Total years smoked
0.0333
10 Y
1.40
0.93,
2.09
Education
0.0226
4Y
1.09
0.59,
2.03
Variable
~
i.2a5a11
~~~
for computations of relative risk. tRelative risk of increase in exposure of one increment, holding other variables in model constant. *Average annual hours in excess of 10 pphm, 1973-1977. gP < .05. llp ,055.
Downloaded by [University of Alberta] at 19:16 30 December 2014
-
concentrations of ambient air pollutants and cancer risks. The study was prospective, and it was based on a population of currently nonsmoking Seventh-day Adventists who were geographically stable for a period of at least 10 y prior to the beginning of follow-up for cancer incidence and mortality. These characteristics of the study population would have removed confounding that resulted from current smoking, and they would have limited any bias that resulted from migration patterns during the follow-up period (either outmigration to escape the relatively polluted areas under study or in-migration). It is possible, however, that individuals who were most sensitive to air pollution moved from higher pollution areas before the 10 y elapsed and were not included in our population. If this is true, the net result would be to bias our results toward the null. All incident cancers were diagnosed beginning in 1977, whereas ambient air pollution concentrations were ascertained beginning in 1966. The high response fraction of 87% made it unlikely that health characteristics that were associated with response were confounded with air pollution. Further assessment of this possibility for which response rates by geographic pollution area are compared is provided by Hodgkin.” We found that response fractions did not differ significantly between geographic areas of study. Many characteristics between high- and low-pollution geographic areas included in the study were compared by Hodgkin.*’ The majority of prior studies have relied upon inexact measures of ambient concentration levels for study subjects. There is an attempt to characterize individual experience in these studies by using census tract level measurements’ and county, and even statewide levels of measurements. The dangers of inferring exposuredisease relationships at the individual level by using aggregate exposure data are well known. We had the same problem to some degree. Analysis of respiratory cancer was plagued by the minimal number of events during the 6-y period of SeptembedOctober 1991 [Vol. 46 (No. 5)]
follow-up (n = 17). We plan to extend the mortality follow-up for an additional 4 y and to conduct a nested case-control study within the cohort with a larger number of respiratory cancer cases and an appropriate reference group. Another limitation of the present study is that ambient concentrations of TSP and 0, were evaluated but indoor concentrations were not. There are no known indoor sources of O,, and the indoor/outdoor concentration ratio for 0, has been estimated to range from 0.1 to 0.7.” It is likely that our population‘s actual exposures to 0, were considerably less than our estimates of ambient concentrations. Therefore, our findings of increased risk are most likely conservative with respect to magnitude of effect. We checked our results by restricting interpolations to be within the A or B quality ranges’ which are considered representative by the Environmental Protection Agency, and the final regression models were run using TSP200 and TSP mean concentration as the exposure variables. The numbers of individuals available for analysis were reduced by approximately one-half and by 6% for TSP and O, respectively. The regression coefficients were essentially unchanged for cancer outcomes after these restriction criteria were applied to both TSP and 0,. However, the adjustment factors are crude and in no way result in a representation of actual human exposures. It is not feasible to monitor human exposures in such a large-scale epidemiologic study during such a protracted period of time or to record the necessary information on personal habits each month for 20 y so that finer adjustments can be made. Indoor sources of TSP were modeled as covariates; the primary indoor source of TSP is tobacco smoke. In the final analysis, these adjustments did not change the health effects results, except to rescale the pollutant. The results of our study suggest that TSP may not be the best measure of toxic particulates suspected of causing adverse health effects. Currently, the fine279
Downloaded by [University of Alberta] at 19:16 30 December 2014
particulate fraction of TSP (< 2.5 p in aerodynamic diameter) is considered the most hazardous. Present plans for our ongoing prospective study call for the estimation of fine-particulate exposures and exposures to nitrogen dioxide, sulfur dioxide, and SO,. Finally, there are problems in separating statistically the effects of air pollutants, whose concentrations in the ambient air are highly correlated. Human chamber studies may untangle this multicollnearity, but such studies are not feasible for free-living populations. The process of relating health effects directly to community concentrations is relevant because air quality standards are based on ambient concentrations. In summary, we observed evidence that supports an association between increasing long-term cumulative levels of TSP and cancer incidence among females in our population. Also, there is suggestive, though not convincing, evidence that TSP is associated with respiratory cancer and that 0, may be related to respiratory cancer.
********** This research was supported by the California Air Resources Board, Grant numbers A6-128-33 and A833-057. Submitted for publication July 1, 1990; revised; accepted for publication April 16, 1991. Requests for reprints should be sent to: Paul K. Mills, Ph.D., Department of Health and Preventive Medicine, School of Medicine, Loma Linda University, Loma Linda, California 92350.
********** References
1. Stocks P, Campbell J. Lung cancer death rates among nonsmokers and pipe and cigarette smokers. An evaluation in relation to air pollution by benzpyrene and other substances. Br Med J 1955; 2~923-39. 2. Buffler P, Cooper S, Stinnett S, et al. Air pollution and lung cancer mortality in Harris County, Texas, 1979-81. Am J Epidemiol 1988; 128:683-99. 3. Euler GL, Abbey DE, Magie AR, Hodgkin JE. Chronic obstructive pulmonary disease symptoms effects of long-term cumulative exposure to ambient levels of total suspended particulates and sulfur dioxide in California Seventh-day Adventist residents. Arch Environ Health 1987; 42:213-22. 4. Detels R, Sayre JW, Coulson AH. The U.C.L.A. population studies of chronic obstructive respiratory disease. Am Rev Respir Dis 1981; 124:673-80. 5. Stocks P. Cancer and bronchitis mortality in relation to atmospheric deposit and smoke. Br Med J 1959; 74-79. 6. Zeidberg L. The Nashville air pollution study v. mortality from diseases of the respiratory system in relation to air pollution. Arch Environ Health 1967; 15:214-24. 7. U.S. Environmental Protection Agency. Air quality criteria for particulate matter and sulfur oxides. Publication no. EPA-600/882-029a. Research Triangle Park, NC: U.S. Environmental Protection Agency, 1982.
280
8. Abbey DE, Moore J, Petersen F, Beeson WL. Estimating cumulative ambient concentrations of air pollutants. Arch Environ Health 1991; 46:281-287. 9. Beeson WL, Fraser GE, Mills PK. Validation of record linkage to two California population-based tumor registries in a cohort study. Proceedings of the 1989 Public Health Conference on Records and Statistics. DHHS Publication no. (PHS)90-1214, 1990; pp. 196-201. 10. Cox DR. Regression models and life tables. J Royal Stat SOC 1972; 34 Ser B:187-220. 11. Dixon WJ, Chief Ed. BMDP statistical software, 1983 printing with additions. Berkeley: University of California Press. 12. Breslow NE, Day NE, Statistical methods in cancer research. Vol II. The design and analysis of cohort studies. IARC Scientific Publications no. 82. Lyon: International Agency for Research on Cancer, 1987; p. 180. 13. Winer AM, Lurmann FW, Coyner LA, Colorne SD, Poe MP. Characterization of air pollution exposures in the California South Coast Air Basin: application of a new regional human exposure (rehex) model. In: Final report, SCAQMD, Statewide Air Pollution Research Center, 1989; pp. 33-47. 14. Phillips RL, Garfinkel L, Kuzma J et al. Mortality among California Seventh-day Adventists for selected cancer sites. J Natl Cancer lnst 1980 65:1097-1107. 15. Beeson WL, Mills PK, Phillips RL, Andress M, Fraser GE. Chronic disease among Seventh-day Adventists, a low risk group. Rationale, methodology and description of the population. Cancer 1989; 64:570-81. 16. Dean G . Lung cancer and bronchitis in Northern Ireland. Br Med J 1966; 1:1506-14. 17. Buell P, Dunn JE, Breslow L. Cancer of the lung and Los Angelestype air pollution. Cancer 1967; 20:2139-47. 18. Hoffman D, Wynder EL. Organic particulate pollutants. In Stern AC, Ed. Air pollution, vol 11, 3d ed. New York: Academic Press, 1977; pp. 361-55. 19. National Research Council. Particulate polycyclic organic matter. Washington, DC: Committee on Medical and Biological Effects of Environmental Pollutants, National Academy of Sciences, 1972. 20. Lave LB, Seskin EP. Air pollution and human health. Baltimore, MD: The Johns Hopkins University Press, 1977; p. 44. 21. Hill AB. The environment and diseases: association and causation. Proc Royal SOC Med (section on Occupational Medicine) 1965; 58:272. 22. Mills CA. Urban air pollution and respiratory diseases. Am J Hygiene 1943; 37:131. 23. Winkelstein W Jr, Kantor S. Prostatic cancer: relationship to suspended particulate air pollution. Arch Environ Health 1969; 18:544. 24. Winkelstein W Jr, Kantor S. Stomach cancer: positive association with suspended particulate air pollution. Arch Environ Health 1969; 18:544. 25. Hagstrum RM, Sprague HA, Landau E. The Nashville air pollution study. VII. Mortality from cancer in relation to air pollution. Arch Environ Health 1967; 15:237. 26. Wynder E, Hoffman D. Tobacco. In Schottenfield D, Fraumeni JF Jr, Eds. Cancer epidemiology and prevention. Philadelphia, PA: W. B. Saunders, 1982. 27. Vena J. Air pollution as a risk factor for lung cancer. Am I Epidemiol 1982; 1 16:42-56. 28. Hodgkin JE, Abbey DE, Euler CL, Magie AR. COPD prevalence in nonsmokers in high and low photochemical pollution areas. Chest 1986; 86(6):830-38. 29. Colome SD, Wilson AL, Becker EW, Cunningham SJ, Baker PE. Analysis of factors associated with indoor residential nitrogen dioxide: multivariate analysis. Indoor Air 1987. Proceedings of the 4th International conference on indoor Air Quality and Climate 1987; 1:405-09.
Archives of Environmental Health
