ML20049J215
| ML20049J215 | |
| Person / Time | |
|---|---|
| Site: | Waterford  |
| Issue date: | 03/08/1982 |
| From: | Campbell V JOINT INTERVENORS - WATERFORD |
| To: | |
| Shared Package | |
| ML20049J211 | List: |
| References | |
| NUDOCS 8203120270 | |
| Download: ML20049J215 (25) | |
Text
..
l a ,
i a
RELATED ColutESPONDENCD ,c . r--
UNITED SI'ATFS OF AMERICA NUCLEAR REGULATORY COMMISSION T2 N.910 P1 :35 BEFORE Tile ATOMIC SAFETY & LICENSING BOARD In the Matter of LOUISIANA POWER & LIGilT CdMPANY Docket No. 50-382 (Waterford Steam Electric Station Unit 3)
SWORN TESTIMONY OF DR. VELMA L. CAMPBELL
- 1. What is your name? By whom are you employed and what position (s) do you hold?
Answer. Velma L. Campbell, M. D. I am engaged in the practice of medicine as the Clinical Director of the Subspecialty Clinie of New Orleans General Ilospital.
- 2. Is this a specialized health field? If so, what is the description of the type of health field?
Answer. I am engaged in Family Practice and consult in the area of Environmental and Occupational Medicine.
- 3. What are your academic qualifications and degrees?
Answer. I graduated from the College of Iluman Medicine at Michigan State University. I completed the Louisiana State University Family Medicine Residency at Charity llospital, New Orleans. During my professional training and career, I undertook special training in the area of occupational and
! 8203120270 820308 PDR ADOCK 05000382 T PDR
i environmental health problems. One of these endeavors was a month at the 9
Occupational Medicine Programs of Cook County llospital and the Great Lakes Center for Occupational llcalth in Chicago, Illinois.
- 4. Ilave you researched the areas of cancer and/or human exposure to toxic chemicals? Please describe your research.
Answer. During 1980,I conducted a review of scientific literature relating Louisiana environmental conditions to human health effects. In addition, I authored an article based on a survey of Family Physicians in Louisiana which described the frequency of occupational and environmental disease in their practices.
- 5. What publications have your works appeared in?
Answer. "The Environment and liuman licalth in Louisiana" was published by the Office of Environmental Affairs, City llall, New Orleans, in December, 1980. " Occupational and Environmental Disease in Family Practice", Velma Campbell, M.D. and Francis I. Nicolle, M.D., The Journal of Family Practice, 13:1, 118-119, 1981.
- 6. IInve you participated in any scientific symposia (colloquia), etc.? If so, when, where, and under whose sponsorship? What topics have you dealt with? llave you held any honorary positions?
Answer. I was a member of the Panel on Occupational and Environmental IIcalth, City Care Conference on the Urban Environment in New Orleans, March, 1981, which was sponsored by the Ecology Center of Louisiana, Inc.; the Urban Studies Department of the University of New Orleans; and the Urban Environ-ment Conference, Inc. I was chairperson of a workshop on Drinking Water Quality at "Towards a Safer Environment: Toxies in the iluman Environment",
January,1081.
- 7. llave you ever appeared as an expert witness in your professional field in state, federal or congressional hearings or courts?
Answer. Yes. Livingston Parish Grand Jury, August,1981.
- 8. Would you please define for purposes of this discussion:
a) Carcinogen u
b) Mortality rate c) Morbidity rate d) Synergism e) Risk of health effect f) llazard to health g) Incidence Answer.
a) Carcinogen - any sustance which can, by itself or in combination with any other agent, cause cancer in animals or humans.
b) Mortality rate - the nubmer of persons in a given population who die of a given cause.
c) Morbidity rate - the number of people in a given population who are afflicted with a given problem, usually assumed to be causing some degree of frank illness or disability.
d) Synergism - the capacity of two (or more) substances whea combined to cause an effect not predicted by the behavior of either substance when acting alone.
c) Risk of adverse health effect occurs when exposure to a given condition (or set of conditions) may cause adverse effects in a population.
f) llazard to health - a condition or set of conditions which are known to have an adverse effect on health.
g) Incidence - the number of cases of a disease in a population over a period of time.
- 9. Would you please describe cumulative effects of toxic environmental chemicals, radiation cc other factors and indicate how this phenomenon (a) would affect health risks to s population? Given Louisiana's high cancer mortality rate and also the known chemical carcinogens present in the environment between New Orleans and Baton Rouge along the Mississippi River industrial corridor, i.e. In Mississippi River are organic chemicals, in the air were found halogenated hydrocarbons, etc., benzene, etc., can you make any statement as to the nature of the risk to the population posed by the introduction of low level radiation (25-75) millirems) that may be introduced into this environment?
Answer. Chemicals, radiation and other agents, when found together in the general environment, may behave in ways not predictable by laboratory exper-iments in which these agents are isolated from each other.
Certain chemicals, particularly halogenated hydrocarbons, accumulate in animal and human tissues overtime, prolonging and increasing the exposure of body tissues to the offending chemicals, thus increasing the possibility of ill effects, including cancer.
One widely held theory regarding the development of cancer is based on the ability of the body's immune system to repair or suppress genetic damage. It i
j is suggested by this theory that there are limits to the repair capabilities of the i
immune system which can be exceeded. At that point, certain genetic damage may go unrepaired and cancer will develop.
Among conditions which may overtax the repair system are aging, stress (e.g. overwork, grief), other illness, and exposure to chemical or radiologic agents in the environment.
L
e f
In the corridor along the Mississippi River between Baton Rouge and New Orleans, Louisiana, there is a larger burden of chemical exposures through air, drinking water and occupation than in many other areas of the country.
The people who live along the lower Mississippi River in southeast Louisiana have rates of cancer which are significantly higher than the national average. Data taken from the SEER program (National Cancer Institution), that is " Cancer Incidence and Mortality in the United States, 1973-1977", shows very high incidence and mortality rates for the New Orleans area compared to the rest of the nation. The average annual age-adjusted incidence rates in the New Orleans area for the respiratory system, including lung, is 71.1 per 100,000, while nationwide it is 52.6 per 100,000. For males, the incidence rate for the respiratory system in the New Orleans area is 128.1 per 100,000. This is the highest in the nation. For bladder cancer, the rate is 27.7 per 100,000 compared to 25.8 in other areas. Females had an incidence rate of 30.0 per 100,000 compared to 24.4 per 100,000 for the total study area. Pancreatic incidence rates for females were 9.3 per 100,000, second only to New Mexico. Incidence rates for white males were 45.4 per 100,000 for bladder cancer. This also is the highest in the study. White females also had the highest at 8.6 per 100,000.
Incidence rates for Blacks showed the highest rates in New Orleans for the digestive system (95.6 per 100,000) and also for the respiratory system (80.1 per 100,000). Black males also had the highest rates for uomach cancer (29.0 per 100,000) and I!!ack females also had the highest rate for stomach cancer (13.3
- per 100,000).
Cancer mortality rates also were very high in the New Orleans area
~
compared to the rest of the nation. The SEER report showed these results for average annual age-adjusted mortality rates: for all cases, both sexes, New l
Orleans had the highest rate for all cancer sites compiled (201.1 per 100,000) compared to the rest of the United States, which was 166.5 per 100,000. New Orleans also had the highest rate for the respiratory system (50.8 per 100,000 compared to 39.7 per 100,000 nationwide); the highest for breast cancer (16.7 per 100,000 compared to 15.1 per 100,000 nationwide) and the highest in the United States in the urinary system (8.6 per 100,000 compared to 7.5 per 100,000). New Orleans was tied for the highest rates with Connecticut for rectal cancer (5.4 per 100,000). Males in New Orleans had the highest rate in the United States for the respiratory systern (94.0 per.100,000); highest in the United States for liver cancer (4.5 per 100,000); and second highest for colon cancer (next to Connecticut) at 23.3 per 100,000. The average rates for females (all races) also showed the highest rates for all sites compiled, at 152.3 per 100,000. Females were second highest in pancreatic cancer, second highest in lung cancer, and second highest in rectal cancer. Females also tied for first place with Detroit for bladder cancer (5.1 per 100,000). Whites overall, both sexes, had the highest rates in bladder cancer in the United States (5.3 per 100,000); in lung cancer (4G.1' per 100,000) and second highest in colon cancer (19.5 per 100,000) and in rectal cancer (4.9 per 100,000). Blacks overall, both sexes, were also highest in the United States for all sites compiled (243.9 per 100,000), in stomach cancer (15.4 per 100,000), in rectal cancer (6.9 per 100,000) and in respiratory system cancer (58.5 per 100,000). This is especially true for respiratory, urinary tract and pancreas. Epidemiologic studies have linked these high cancer rates to such expmures as use of the Mississippi Itiver for drinking water, employment in shipbuilding and chemicalindustries, and residence near petroleum refineries.
A study done on lung cancer in Louisiana through death certificate analysis (Gootlieb, Pickle, Blot, and Fraumeni, Jr., JNCI, November, 1979) revealed l
} ,
approximately a twofold excess risk of cancer associated with certain types of industries, such as shipbulding. Lung cancer risk was also found among older men who had been employed in the petroleum industry and among male and female residents of towns where the petroleum industry was a major employer (see i
enclosed article).
In a study on pancreatic cancer mortality in Louisiana (Pickle, Gootlieb, 4
AJPil March,1980), high pancreas cancer mortality among white males in a u
cluster of Louisiana parishes was investigated. Excess risk was seen for workers in the oil refining and paper manufacturing industries, and for residents living near refineries.
The latest Louisiana study on " Cancer and Drinking Water in Louisiana:
Colon and Rectum" was published this year (Gootlieb, Carr, Morris, IJ of E,1981, see enclosed). This study found a significant risk for rectal cancer associated with drinking water derived from the Mississippi River. This was a case-control study conducted in twenty parishes (counties) in South Louisiana. The j association between the drinking water and the cancers was not dependent on l age, race, sex or year of death. Chlorination also associated significantly with I
rectal cancer. Among those who used river water, the risk increased inversely as the distance from the mouth, with greater risk downstream from the many industries which line the river.
The differences between Louisiana and the rest of the United States in such things as dictary patterns or cigarette smoking do not account for such a large difference in cancer risk.
i Although data for areas other than Metropolitan New Orleans regarding cancer incidence and mortality are limited, national statistics indicate that, of
! the 18 countics in the United States with the highest cancer mortality rates,11
I are Louisiana parishes between 11aton llouge and New Orleans along the Mississippi itiver.
The peopic of this area, then, face a potentially serious public health
~
problem. They are exposed to a vast array of chemicals from a variety of sources. They also suffer a burden of cancer incidence greater than the national average, whleh are demonstrably related to those environmental exposures.
Now, it is proposed to add another increment of risk to the already higher than average burden. Although it is not possible to quanitfy the level of additional risk, the medical litera ture includes significant research which supports the premise tint small doses of radiation increase the development of cancer from exposure to some chemicals. Therefore, the logical conclusion is that to knowingly add radiation, even at low levels, to the chemical exposures confronting the presumably limited capacity of the human immune system is to greatly inercase the risk of cancer for each individual who lives m the area.
- 10. Ilow would the presence (action) of carcinogenesis be manifested in the population? Could this health risk be cumulative over the lifetime of the individual?
Answer. It would be manifested by increased cases of cancer in the population. Yes, this could be cumulative over the lifetime of the individual, if the individual continues to be exposed to carcinogenic substances.
- 11. Could you indicate any categories of individuals more likely than the rest of the population to demonstrate health c fects of toxic substances?
Answer. Small children (less than seven years old) and older people (sixty years and older) are particularly vulnerable to this type of risk, depending on the type of substance and the condition under discussion.
- l. .
.- A N. ,
, *( ... ' 'p)_
]O Ol C )
- :niernational Journat of Ep.de niology . Vol.10. No. 2 p
- ginted in Great Britain 3 Onford Uneversity Press 1981 e
, 3
- 2. . .
Cancer and Drinking Water in Louisiana:
p~
Colon and Rectum MARISE S GOTTLIEB,' JEAN K:CARR** end DANIEL T, MORRIS **
Gottlieb MS (Department of Medicme, and Department of Biostatistics and Epidemiology, Tulane University School of Medione,1430 Tulane Avenue, New Orleans, LA 70112. USA), Carr JK end Morris DT. Cancer and drinking weter 6n Louisiana: colon and rectum. laternationalJournaf of Epidemiology 1981,10:117-125.
A case <ontrol mortality study concbeted in 20 parishes (counties) of South Louisiane to determine what relation.
j/* ship drinking Mississippi River water might have on mortality from colon or rectal cancer, found a significant risk for rectal cancer associeted with surf ace water. Rectal and colon concer deaths E92 end 1167) from 1969 to 1975 were matched to norKancer deaths by age et death lt5 ysers), year of death,een and race, and withir. groups of perishes with simiter industrial and urtaan-rural characteristics, each group being defined so as to ensure that it included as neerty as possible equeLpopulancos.Arsing yveterJronqgg,n taruisur_fpce_sou_rces, based on the 1970 census. Colon
(
s cancer did not relate signittcantly to any water venetde, but rectal cancer associated strongly with surface, or Massassippi River, water. The odds ratio for rectal concer between those who were bom and died on ground water
- was 2.07 with 95% C.I.:(1.49-2.88). A multi <$imensional contingency table analysis found the association between rectal cancer and surface water significant at the .0001 level and not dependent on ope, race, sem or year of doeth.
- #w The risk:for men was slightly higher.then for women,-but both.semes showediserincreased ~ - risk. Chforinetionh .
associated significantly with rectal cancer. Among' 5ose who used river water,'the risk increased inversely as '.hei -
distance from the mouth, with greater risk downstroom from the many industries which line the river.
- f. The presence of carcinogenic substances in public cultural, ethnic and dietary differences between drinking water has recently led to concern about surface and ground water parishes which may be possible relationships between water quality and related to the increased rate of cancer portality.'
g This point is panicularly relevant to the study of
. high cancer rates. Studies of surface water use and
, overall cancer mortality have shown associations in colo-rectal cancer because of the role diet is believed Louisiana,s.2 Ohio' and Missouri.* Since Louisiana - to play in its actiolop. The association of colon has some of the highest mortality rates of several cancer with high fat' and low fibre" diets suggests sites of cancer in the U.S.5 and Mississippi River that both factors are likely to be important in e,olon water, which is used by many 1.ouisiana pa-ishes as cancer actiology, while _ differences in dietary their source of drinking water, has a large number of patterns are known to exist in Louisiana, 2 ,,
' organic chemicals," Louisiana was a natural loca. dietary study has been done using a questionnaire tion to investigate this association. Using aggregate comparable to those used in other colo-rectal cancer I statistics, both ihrris et a1 8and DeRouen and studies. Therefore the ' impact of these dietary differences on colon cancer mortality can only be b Diem' found an increased risk of colo-rectal cancer associated with surface, or Mininippi River.wajter spenlated. . .
use. ....~ U. Due to the association between colo-rectal cancer
}. and surface water use shown by aggregate studies, l' One of the difficulties in interpreting studies using aggregate data from Louitinna is that there are several case-control mortality studies have been l 83 pursued. In New York State Alavanja found signifi-cantly elevated odds ratios for both colon and h
g .l
{
- Department of Medicine and Department of Bio-statistia saf Epidemiology, Tulane University School " Medicine, 1430 Tulane Avenue, New rectum in males associated with chlorinated water use. A case-control study in North Carolina s4 has Orleans, na 70112. USA.
also been undertaken, but the results from this
- Health Services Research Institute.Tulane University study have not Iet been reported.
p,e ,
I School of Public Health and Tropical Mediesne New The use of inc.dence rates rather than mortality i
l oric ns, usa, l Repriers from Dr M S Corsheb. rates has shown associations between levels of l 117 .
l l
l
. ? \
1 118 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY ,
l ' death from residual are used to approximate risk factors during organics in drinking water and oesophageal and pan. . remainmg [
creatic cancer, though not colo rectal cancer, in the time preceding death, and water soure at birth controlied is used to estimate length of exposure. De cas'es New York State.'8A more detailed discussion of the and controls have been selected to minimise dieury, ,
from this - ,
strengths and weaknesses of the evidence for cancer 1 Mi551552PP5 and drinking water may be found in a recent review urban and industrialisation biases. Colon and rectal ' The par article.8' cancer will be shown to behave as different similar ind neoplasms, with pos:ibly different actiologies. In.
One difficulty with studies based on the resi- l and each gi dence at death, however, is that this residcu may dustrial exposure by occupation and residence have also been controlled and examined, included e not in fact represent the residence during the time using water ci exposure to the carcinogen, which is likely to on the 191 have occurred atleast 10 years earher.This problerra MATERIALS AND METHODS was randorI b particularly important for colo-rectal cancer The 20 parishes (counties) selected for study, shown race, sex c because of the relatively long survival time between in Figure 1, were chosen from the southern part of ,
from withe diagnosis and death, during which the patient could leuisiana, where cultural and dietary habiu are and contrc move to an urban, surface water source which may relatively uniform. In these parishes between 1960 characteg be closer to the availability of long-term medical and 1975 there were 4723 colo-reetal cancer deaths confound r treatment nig could create a spurious association list.d in state vital statistics files: 1141 rectum ditions, we between the cancer risk and the water source. De (ICDA 7,8:154) and 3582 colon (ICDA 7,8:153).
Control de case-control studies mendoned above deal with the A combination of systematic sampling procedures cells const lifetime water use problem by restricting the popu- resulted in a study population of 692 rectum and
! race, parisl lation to those living in counties with low migration, 1167 colon cancer deaths. This desired number was determined from the relative odds observed in other_,____ _ method og but did not validate length of residence on an elsewhere.
studies. For rectum the sampling consisted of every
. individus1 basis. C -To-desc 7 ' CIn 'this-itsdyMliaracteristics~oTGieinsent/W"thfid"dTal'iroizfUrTeinF;ilrish 'and ail deaths'from>~- and utilisei dividuals including loss on ignition and chlorine the remaanmg parishes and for colon every sixth death cer
^ " "" 5^ 5 reviewed i Characteri:
ignition a1
- l. company I control sa first billin, directorici ascertain 1 at the res' related t3
.outstama e se ets s the water i areas not mississtrra ground w trus ___
source, wi
.- *
- Each y
. . a *** within 0.2
.**.=l*..l. . ' '
- l* **** ordinates
- e. . .
i.
- l.
rural add 1 mation w.
v ,
- .j -
~y Using (1942-11 was iden' cutf of unsic. tured by g Ground Water supplied '
' Industrie,
@ Surface Water this locat Mixed cround and '
the 4 ma Surface Water ,
FGURL1 Map of Louisiana whb location of study parkbes
i . \ .
l.
I ,
x I
119 BOWEL CANCER AND DRINKING WATER I
recorded on the death certificate was coded using
>rs du.mg i death from Orleans and half of the deaths of the remaining parishes. The lighter sampling of Orleans the U.S. Census Code of occupations and industries.
' e at birth This enabled the analysis of confounding by l The c: es controlled for a bias due to an imbalance of cases from this single densely populated area using the industry of occupation and industry of residence e dieury, Minissippi River for iu source of water. from effluents in local ambient air in relation to a md rectal water source effect. Timespace clustering of cases,
' different he Parishes were clustered into groups with sirrilst industrial and urban / rural characteristics, associated with local water run offs from industrial
- ogies. In. effluents,if any, could also be detected.
I ence have and each group being defined so as to ensure that it included as nearly as poulble equal populations One of the unique characteristics of Southern using water from ground and surface sources, based Louisiana is the presence of Acadians. The high on the 1970 census. Each colorectal cancer death canect rates in Southern levisiana may be due in was randomly matched by age 'at death 25 yean, part to some genetic or dietary factor associated
. ly, shown with the Acadians, who are descendants of the l n part of race, sex and year of death to a non-cancer death from within the same parish group. Therefore cases French speaking inhabitants of Nova Scotia. ney habits are and controls were similar in industrial or urban have largely maintained a unique culture and l cen 1960 customs, and are relatively inbred. Based on a detail-
- er de:ths characteristics but not in water source. Deaths from confounding causes, i.e. other gastrointestinal con- ed study of the original ships' logs and tracings of J I rectum modern spellings a list of Acadian names and a ditions, were unavailable for the selection of controls.
, g; 3 3 3),
control deaths were randomly selected from within coding scheme has been developed which allows the roc:dures Acadian ancestry of each person to be coded.
cells constructed from 5 year age groups, by sex, crum and Since the estimated lateney period for cancer is mber was race, parish groups and year of death. The specific method of computerised control selection is reported likely to be at least to yean, the water source at J in ether death is not necessarily the water source ofinterest I cf every elsewhere." c
- - < - %To determine theyean cach person had lived ats when , investigating..the.,actiol.ogy of_can. er. To , _
aths from and utilised water sasociated with the address o~n the estimate lifetime kater use for those with unknown" --
ery sixs water source 10 or more years before death, the '
death certificate, water company records were reviewed for the earliest billing date to the family. following gradient us constructed to reflect the Characteristics of the water including loss on amount of lifetime drinking of surface water.
ignition and chlorine residual were obtained from 'mostly surface' (those who were born and died in a company records of routinely collected quality surface water parish),'some surface'(those who had control samples for the yean between the person's some known surface water use, either at birth or first billing date and year of death. In addition, city death),'possible surface' (those who died in a directories and voter's registration were used to ground water pa ish but had either unknown or out ascertain the earliest known date the penon resided of state birthplaces), and 'Icast surface' (those who at the residence at death, and this information was were born and died in a ground water parish). While related to information on the dates of operation of water use between birth and death is not known, the water companies and the areas served. Rural these categories reflect a gradient of surface water areas not served by public services, where shallow use (an attribute for which the term "sourcelife'is used hereafter) such that the mostly surface group ground water wells were the only available water source, were coded as such. would contain most of the long surface water users, Each penon's residential address was located to while the least surface group contains fewest of the within 0.2 of a mile, using latitude and longitude co- long surface water usen. In addition to this L ordinates 'on the map. Iecation of subjects with a gradient, the additional information on' length of
" rural address or incomplete death certificate infor- residence collected in the field provided detail for
[ matmn was facilitated by contactmg local agencies. individuals whose length of residence was known l and was longer than 10 yean.
l Using The Louisiana Directory of Manufacturers (1942-1975), each industry in the study parishes ne primary method of analysis used is the was identified, characterised by product manufac- multidimensional contingency tabic analysis, with 6mssed by convenion of the Standard Industrial Code selection of an optimal model as described by supplied by the industry to the U.S. Census Code of Brown.sa which is based on the log-linear model and l Industries and located by map coordinates. Using utilises maximum likelihood estimations of all this location each person was given a distance from possible first and higher order effects, (BMD the 4 major industries closest to his or her residence package program P3F). The variables used to con-at death. In addition, the industry and occupation struct the models are the matching variables. age
m p
120 INTERN ATION AL JOURN AL OF EPIDEMIOLOGY (dichotomised at the mem); year of Jesth of relatedness of exposure and disease, and upper i .
(1960- 68 or 1969-1975, the midpoint of the and lower bounds of the 95% confidence interval
_. study period by number of cases); sex and race.The were determined by exact methods outlined by analysis was then run 3 times for each cancer site, Gart.8' Cluttering in time and space was examined using 3 different estimators of exposure Sourcelife using the procedure designed by Mantel,'D and (described above); chlorine level (none, low fless performed separately for cases and controls.
thrn 1.09 ppm) or high (greater than 1.08 ppml) and loss on ignition (Iow [lcss than 68 ppm] or high RESULTS '
[ greater than or equal to 68 ppml). Both the ne deaths due to rectal and colon cancer were chlorine level and loss on ignition were divided at matched to equal numbers of controls resulting in a the mean. The recommended chlorination lew! for rotal population of 3718. As presented in Table 1 fmished water was 0.2 to 0.4 ppm for the period bf the rectal cancer deaths were 70% white and 54%
data collection. The loss on ignition value, when male while the colon cancer deaths were 75 % white ave.ilable, approximates the level of orNanics in the and 46% male.
drinking water. A model is fit which best describes Of the deaths, 3243 or 87.2% had residences the data structure for each exposure variable, and which were able to be located to within 0.2 of a which rari disease. The closeness of the fit of the model is mile and another 115 located to within 3 miles. of 1.61 ((
measured by a likelihood ratio chisquare, as is the Field validation of length of residence was less com- the water significance of the contribution of each of the terms plete 51% of the sample, or 1886 people, were treater th cf the model. either missing length of residence or had a length of lationship To assess the possible impact of accumulation of less than-10 years, which meant that water source variable affluents along the river among those drinking during the probable carcinogenic period was un- associatio Minu<iippi River water,~the distribution of cases and , known. As presented in Table.1,3 total of 3688 or --
-scen,in t)
.. controls who died on surface water was exammed 99.2% were ~
able~ to be classified using the ' source. .four carc
, 2~ I by theirWasoi along the livcr'~For*convenie~nced FEfiaile.'InTe li cMti'n~g'esiWtan shalysis usin'g^~~ MMern this was trichotomised into those who lived above chlorine as an exposure variable 2932, or 78.9%, indicates Orleans, in Orleans, or below Orleans parish. had a chlorine value and were able to be used in the steadily a Residential proximity to industry and industry analysis. Loss on ignition was available for 1864, or increases <
ef occupation were also examined for each site to 50.1 % of the total population. least, is 2 detect associations with excess risk. De effect of Since the results for the 2 sites of cancer differed , to 2.88.
Acadian ancestry was determined using odds ratios substantially from one another their results will be The m and confidence intervals. The length of residence presented separately, resulted i collected in the field was examined in a categorised model foi manner to confirm the results of the sourcelife vari- a) Rectum for 99%
rble. . Using water source at death, there was a significant (p=. 000 Odds ratios were calculated to express the degree risk for rectal rancer associated with surface water, and wate matched TABLE 1 DescriptiN sunristics of rectal and colon concer study popadation g Cancer of the flect thei
. Rectum Colon ne fit G firming t Total population .m i ent gmm (cases and controls) 1384 '
M*? . 2334 chisquare a o. . rx . . . . .N-- ., .
Race (% white) 70% , ; ,, ,
75% ,
The m Sex (E male) 54 % s: , 46% ~ showed &
-66
~
P 67 the exp l Age (mean)
Controls (P=0.12).
Numbers used la Cases Controls Cases surface w l Sourcelife analysis - 689 686 1161 1152
% :otal 99.6% 99.1% 99.5% 98.7 % that dise Chlorine analysis 546 5 14 951 901 77.2% 31J% 77.2 % level was
% total 78.9% -
for whici Loss on ignition 142 ~ 517 635 570 -
% total 49.4 % -
45.8% 54.4% 48Js And IT0" controllir 1
~
I I2I BOWEL CANCER AND DRINKING WATER TABLE 2 Assa of ressel n=4 telen enacer eneraalisy sisecissed irish leng 6 of espesvre se i ed upper d "* * '"E * * '/'" 'd ""* * *" /' !
interval ined by Cancer of the xamined Rectum Colon odds odds J, and ratio Cases Controls rata Sourcelsfe Cases Controls Mostly eface 286 219 247 370 380 0.96 Some surface 218 214 1.61 381 375 1A1
,7 ,,7, Possble marface 96 112 1.36 208 197 1.05 ting in a Table 1. Least surface 89 141 202 200 f T
bad 54% 0.96 Most to least 2.07
% whitz (95 % conradence limits) (1.49 -2.88) (0.75-1.24) pidences the surface-ground effect on rbame was of border- I D.2 of a which ranged from an odds ratio (surface to ground)
D miles. of 1.61 (0.91-2.85)in those with 10 to 19 years on line significance (p=.05), implying a chlorine effect ess com- the water source to 2.11 (1.17-3.84) in those with in addition to the water type.
greater than 30 years on the source. % hen this re- De results of the loss on ignition analysis are /
fe, were '
yngth cf lationship was examined using the soureclife presented in Table 3. The model shows a significant var:able to approximate lifetime water use, the term which is a four-way interaction between ysource was vn- association was further strengthened (Table 2) as disease, loss on ignition, age. and year of death.
D688cr seen in the distribution of cases and controls by the . Overall a loss on ignition value was available for 659
~~
~sodree; _ "'fo~iIr~ cat'egoriisTtliis' variable.'ne' odds .rztios *~ pe6plegr"48%"o7theyp'ulition:HowcWfTsrsIller_
6 Esing between each category and the least surface category proportion (40%) of those dying before 1969 had 78.?%, indicates that the risk for rectal cancer increases values for loss on ignition, while the proportion of ld in the L
steadily as the estimated amount of surface water the later deaths was 60%, with proportionally more low values in the later years of death (Table 4).
l864, or increases, and between the 2 cxtremes, most so least, is 2.07 with a 95 % confidence interval of 1.49 There is a direct association between rectal cancer differed to 2.88. and high loss on ignition in 2 of the 4 groups the I will be ne multidimensional contingency table analysis older people dying before 1969 and the younger j resulted in the models presented in Table 3. He people dying after 1968, but only in the latter model for the sourcelife variable, which was present group is the association significant (p=.04). He for 99% of the population, shows a significant odds ratio of high to low loss on ignition in this i
nificant (p= 0001) first order interaction between disease group is 2.53.
water, and water source. Since the cases and controls are in order to determine whether any industries of matched for all the variables other than water occupation were assocsated with an unexpectedly source, the other interactions seen in the model re- large number of cases, they were aggregated into 22 flect their interrelatedness independent of disease, groups. Although several of these groups had more The fit of the model u a whole was excellent, con- cases than controls, the small number of people in firming that the model was not significantly differ. the industry group prevented them from achieving ent frow tKe data? ,with"the significance of the. significance. Thie construction industry, however.
. chisquarc'almost eqsial'to 1.0. & had 44' cares and 24 controls and this difference was ne model for chlorine also fits well(p= 82) and significant at the 0.05 level.The odds ratio of those
, showed a significant first order interaction between working in the construction industry to those work-the exposure vanable, chlorine, and disease ing out of it was 1.78 (1.07-2.96). The risk was (p-o.12). Since chlorine is used in higher levels for not due to confounding of surface water with the surface uter than for ground water it is conWent construction industry, but was found to exist that disease, which associated with surface .rer, mainly in 2 parishes Orleans and Jeffenon, where also associates with chlorine level. lew chlorine the number of cases was 3 times the number of level was the only level of chlorine (no, low or high) controls. When the occupations which made up the for which there were adequate numbers of surface construction group were examined, the risk was and ground water to compare the 2 water sources, found to exist largely among carpenters and controlling for chlorine level. when this was done electricians. Because the numbers of individuals
I l i 122 INTERN ATION AL JOURN AL OF EPIDEMIOLOGY TABLE 3 Goodness of fit paramasers of optimal modelfor source life, chlorme level and loss 4n ignition multidamensional contagency sable analysu
~ CANCER OF THE RECTUM Model 1sems item LRx*' p Laposure variable - LRx** p d f.
dJ.
Sourecitic Source-sca-race- .0044 3 13.13 year of death Age-sez race- .0208 1 5.35 ycar of dcath 3 .20J9 .0001 Disease-source 3 930 A204 Sourcesse 43.86 .9998 81 OVERALL MODEL where 1 Chlorine-age sen-Chlorine other c race-year of .0359 was tra; 2 5.77 death 2 8.86 .0119 in bus Disease chlorine .8232 45 36.18 portati4 OVERALL MODEL ever, u Loss on ignition Discase 40ss on ~ but am ignition-age I 6.16 .0131 year of death
,8.43 . 0o37 Z :' C
- _,- -- woskei 1 _
. Age-ecx
.,.m - -. sg. -+ .a 7.it_Jtati
- ' 1?! Servace ; 4J:21,, , g 5.0 y6.: :~g 0243 3,_ y .-.- -. ~- that 11
'YI' _"',1 ,
44 29.27 .9570 ,
cultitri OVERALL MODEL Clu
%aimum likelihood ratio chisquare (partial association between f actors) cances a miles h) Colon Cancer moss involved in these risks is too small to disturb any As apparent from Table 2, the distribution of colon overall results, no correction for occupation was associ cancer was not related to the sourcelife variable or close necessary in the water analysis. Mhen residential to any other water variable. This was reflected in proximity to industry was examined none achievedthe model with the best fit from the contingenry Livig significance, although it is interesting that there was a table analysis, presented in Table 6.ne significance were 43 cases compared to 30 controls who lived withi of the chisquare was extremely close to l.0(p=.996).
within one mile of a furniture manufacturing proni which meant that the fit of the model was excellent. - A.
company. No association was observed with Acadian but disease did not relate significantly to any other ancestry. The rectal cancer cases and controls were variable in the model, and in fact makes no signifi- of cc examined separately for time-space clustering based cant contribution to the fit of the model (p=.85).
on residence at death and no clusters - . .
were observed The implication of this is that the cases and controls for either cases or controls.- were similarly distributed throughout the multi-
% hen surface uter users, cases and controls, are dimensional contingency table that was analysed.
examined by ;4 heir , location along the Mississippi _ The models for chlorine and for loss on ignition also River, there.are more cases down river and more, contained disease as a lone variab!c, not contribut-controls .cp .siver from .,New Orleans. The l
ing anything significant to the model. As shown in distribution of the cases and controls is presented ira Table 5, the location along the river also had no im-i Table 5, with the odds ratios between categories pact on disease among the surface water users.
l Due to the overall association of rectal cancer when colon cancer was exammed to determine ~
- with surface water, there are more cases than whether' there was any relationship between occupa-erantrols in the total group of surface water users, tion and cancer, no occupational groups were signif-This imbalance, however, is not observed in the icant at the 0.05 level. Two groups, however, were group above Oricans, where controls predominate, close to significance and were examined further. .
slightly. The odds ratio of those belv n Orleans to Finance had 125 cases and 97 controls, with the those above is 1.82 with 95% confide .cc intesval of largest difference found in administration, .
1.01 to 3.26. I i
r
=
123 BOWEL CANCER AND DRINKING W ATER TABLE 4 Risk of cancer of sbe recrum morsality associated ui 6 loss on igwision of drraking
,,,,,,,,g unser age and yens of drash CANCER OF THE RECTUM Year of death 1960 - 1968 1969 - 1975 P
Age A8e
<67 years 67+ years <67 years 67+ years Loss on ignition Controls Ca Co Ca Co Ca Co Cases 28 41 36 24 9 17 24 High ( ) 68 ppm) 28 38 43 60 57 81 85 Low (< 68 ppm) 53 35 0.66 1.29 233 0.74 Odds Ratio (High to Low)
.9998 where there were 39 cases and 21 controls. The DISCtissION One of the first concerns when conducting a study other occupational group which was investigated was transportation, and the risk was found to exist of drinking water is to avoid confounding dnnkmg in bus drivers and in those engaged in water trans- water with other factors in the environment which J232 portation. The risk for water transportation, hoe might relate to the disease under study. To avoid ever, was not found among sailors and deckhands this bias the parishes were grouped according to but among longshoremen and clerical employees of industrial, urban characteristics and matching of cases and controls was performed within the parish shipping companies. While the risk for bus drivers,
._ -tworkers in-water. transportation and administrations group @su,ccess Of,,the_ design in accomplishing ~
is staris'ticaDy significant, internal eviden'ce sQigests t his'bbjective is shontiy~'tEmbseiicTofiny rasjoi,.
that it is likely to be a sociotconomic, dietary or effect of Acadian ancestry, industry of occupation
- C570 g or proximity of residence to industry.The fact that cultural effect rather than occupational.
Ouster analysis detected no clusters of colon colon cancer, a condition which is strongly asso-cancer in space or time. However, residing within 3 ciated with lifesgle factors such as diet and socio-l miles of a moss processing plant, where spanish economic status : .21 and which was present in the t
of col:n moss is processed into furniture stuffing, was majority of cases in this study, shows no association riablz or associated with risk of colon cancer, with 16 cases with water source is further validation of the close to the plant compared to 5 controls (p=.03). success of the study design in preventing the con-hedin Living within a mile of furniture and textile industries founding of urban lifestyle with surface water. The PngeneY findings of aggregate studies'# demonstrating hific*nce was also associated with risk at the .05 level, and p=.996). within a mBe of petroleum industries with a less associations between colon cancer and water source may be due in part to this type of confounding.
feeDent, pronounced risk, (p=.08).
Acadian ancestry showed no association with risk %c large difference between colon and rectal py cther cancer in relation to drinking water in these results o si of colon cancer. -
,(p=gmfi-
- .85).
,contr:Is u ,
,'i u:T TABLE 5 . Odds meio for cancer of the rectum sad colon mortality associased wish location f multi- i e 3g . ofsurface untersource on sheMississippirieer, _ 3 ,_
malysed.^
tion also ' .. . ,
- Cancer of the Cancer of the yg p .,,';
- , . recrum colon bzwn in ~
Mississippi River Odds location of wster Odds ,
2noim-s y t'a = Controls ratio Cases Controls ratio source k
"' 76 50 1.82 82 75 0.91 Below Orlcans l occupa- 276 1.41 498 511 0.81 Orleans 325 re signif. 76 63 Above Orleans 41 49 h, were 1.82 0.91 l further. Below to abovei Wh the ,
(1.01-3.26) (0.56-1.47) kstrati:;n, (95% Confidence Limits)
e . .
. I !
a
- 4 a 124 INTERN ATIONAL JOURNAL OF EPIDEMIOLOGY l , .the true ino TABLE 6 Goodness of fit parameters of optrmal modelfor sourcehfe esposure using multi- carcmogen demensional contingency sable analysis - restrict the urgent nee.
CANCER OF THE COLON Model Item residents m Esposure variable Model d.f. LR z p dl LR m' ' p chlorinatec Althoug Sourcelife Age-sex-race- further sic
! 8.56 .003 year of death '
water risk 3 24.02 .o00 Sourcerace vestigation.
Source-year A case-s
,3 17.18 .002 of death .
.m th.u an 1 0.03 .85 carcinogcE Disease 302 67.92 .996 informatis OVERALL MODEL cancer in L
- naamimum likelihood ratio chi square Ipartial annociation between factors!
ACKNOW, proportion of people who were found and in the Protection is striking since the 2 sites were treated identically bias observed with year of death. The procedure wish to t fgr control seicetion and data collection. The that water companies used in co'.lecting loss on """**"'*"
models for both car.cer sites fitted the data almost ignition evidently changed during the period of the exactly, but colon' cancer related to no other study (1960 to 1975) such that the chance of find-variable in the model, while rectal cancer related -
ing a loss on ignitiori'value'.for a person was snuch,-
strongly and consistently to surface water and . . , - . , , , , , , ~,
- .chlorinationdvidenceof epidemiological.differenges,bettet .in the.later._y. cars than in the_carliegears.
3 Harris I
' betweenand ~
iheis cancers'of these sitesyears strongly confirmed in these h'as been might be duenoted---ibe trendinobserved either to an improvement the towards Mi'5i lo before,22- organics level of the water or to collection of the Funs results, including some of the industry of occup ' 2 data more often when the level was low. Given these canc ation associations noted for colon cancer which problems it is advisable to consider the.results for 3 suggest other cultural or dietary associations. The suncher the earlier years as more suggestive than conclusive. lask association of drinking water with rectal cancer For the later years, however, close to 60% of the rather than colon cancer is also physiologically population had a loss on ignition value, and the ob-4
"{
g;,,
reasonable since the function of the rectum is water served risk ratio of over 2.0 among the younger resorption and the faecal material remams a longer 196<
time in the rectum while it travels rapidly through people may represent a real risk for high organics. ,SEERI He fact that the risk is fr.und in those dying U the colon. younger than the mean age is particularly disturb- Ed ne association of rectal cancer with surface ing in terms of future expected rates of rectal cancer.
is-water and chlorination raises the question of what
- oo.s>
The association seen between rectal cancer and drin component of the water is initiating or promoting location along the Mississippi River, with increasing "**
the cancer. The high levels of chlorinated hydrocar-risk as one lives closer to the mouth of the river, bons found in the Mississippi River drinking water" 7 I
many of which are known carcinogens, are likely further confirms the overall association of rectal gd',', rep
! cancer with chlorinated surface water, ne highly Cin, candidates for actiologic agent. If this hypothesis developed industnal nature of the river, with 197 were true then the disease should correlate with occasional spills from large ships and dumping from ,"***
' chlorine level and with loss on ignition, a men'sure of crganics in the water. Since all surface water is chlorinated it is impossible to determine the effect major industries along the river, make it likely that carcinogens find their way into the drinking water h Vol of Southern Louisiana, with increasing concentr- Hai cf high chlorine without surface source. The fact ,De Re ations as one travels downstream.
_ that the surface . water effect, however, is of only ne results from this study suggest that at least borderline significance when chlorine level is con-
~
. one cancer site is strongly linked to the consumption trolled ~ nplies that chlorine does make a contri- of this surface water, bued on mortality alone.
bution to the risk for rectal cancer. Since mortality rates are known to underestunate ,
The results from the loss on ignition analysis are risk, it is probable that the relative risk would be more difficult to interpret. Field collection of the even more pronounced than these results suggest if loss on ignition value was problematic both in the
1 T
BOWEL CANCER AND DRINKING WATER 125 the true incidence were known. Identification of the Cancer Mortality in Louisiana Drinking water Source
' and Other Possible Causative Agents. Origins of carcinogen (s) responsible for the risk in order to Human Canca, Han, wauon. Wmsten. Volume 4 restrict the level of them in drinking water is an Book A. pp 331-345, Cold Spring Harbor Lab-urgent need not only for this area, but for all urban oratory,1977.
residents world wide who drink water derived from IO Wynder ER and Bardaru SR. Dietary fat and colon chlorinated surface sources. I Burkittcancer.J Natt Cancer last 1975:54 7-10.
Although this study does not show causality, it DP. Large bowel cancer: An epidemiologic jig-saw puule.J Natt Cancer inrr 1975 54, 3-6.
further strengthens the credibility of a drinka.ng 12 Steciman VP. The cultural context of food. Center for water risk and provides evidence to further the in- Agricultural Sciences and Rural Development, vestigation. Bulletin No. 681, Louisiana State University, Base,n Rouge,14uisiana,1974.
. A case-control interview study on incident cases 33 Alavanja M. Goldstein I and Sumer M. A Case-Control in th.is area with concomitant water testing for Study of Gastro Intestinal and Urinary Tract Cancer j carcinogenic agents may provtdc sign 2ficantly useful Mortality and Drinking Water Chlorination. Water information for the prevention and control of rectal Chlorination: Environmental Impact and Health
! cancer in Louisiana and elsewhere. Effects. Vol. 2. Ed: jolley RL, Gorchev H and Hamilton DH Jr. Ann Arbor Science Pub. Inc.,
, ACKNOWLEDGEMENTS 34 P*3' N Scruba RJ and'IU8-Shy CM. Cancer and Drtalung Wate r .
This study was supported by Environmental Quality in North Carolina: A CaseControl Approach in the Protection Agency Grant No. R 805 110. We also Utilizing Prior W ater Uw Exposure Gradients
>ocedure oss on wish to thank Dr F Mather for her advice and (abstuct) Program and Abstracts: frore American assistance. Public Health Amociation,107th Annual Meeting,
% of the N.Y. N.Y., p. 201,1979.
is
)cf find- Carlo GL and Mettiin CJ. Cancer incidence and tri-much . halomethane_ concentrations in a public drinking
. - , - - - . . - .z watersystem. A m j Pnebhc Nulsb l 950.70d1J.-425.e -
- years. -. . , . e-- -
a - 16. .Wilkins JR III Reiches NA and'Kruse CWl Organic
~ '
@clater REFERENCES chemical contaminants in drinking water and cancer.
.t in the Harris RH. Implications of cancer causing substances in Am J Epidemiol 1979 110, 420-448.
37 g.cg,er F Scale D and Gottlieb MS. Random selection cf the Mi'*i'SiPPi River water. Environrnental Defense Fund.1974, of matched cases and controls from large ordered
,en 4 ese 2 Page T. Harris RH and Epstein SS. Drinking water and computer files. (Sub'mitted for publication).
Ilts f:r cancer mortaliryinleviaimaa Scitare 1976;193 55. 38 Brown MB. Screening effects in multidimensional con-
)clusive. Buncher CR. Cincinnati drinking water - an epidemio- tingency tables. Applist Srsristics 1976 25 37-46.
logic study of cancer rates. Cincinnati Ohio: 19 3 cf the Gart JJ. The comparison of proportions a review of University of Cincinnati Medical Center,1975. significance tests, confidence intervals and adjust-
) the ob- 4 Cook GB and Watson FR. The geographic locations of ments for stratification. Arver del *lestirare laste-pounger Missourians with multiple cancer. Missowri Medicist national de Saaristiget 1971 39:148-169.
prgarucs. 1966;32: 997. 20 Mantel N. The detection of diseau clustering and a 3 dying , SEER Programi Cancer Incidence and Mortality to the generalized regression approach. Concer Ras 1967 United States 1973-1976. Department of Health. 27, 209-220 dht 6 Education, and welfare. Publication No (NIH) 21
' CTDCCT- Hoover RT, Mason TJ, McKay FW and FraumenijF Jr.
18 37 Bethesda, Maryla,d. Geographic Patterns of Cancer Mortality in the U.S.
icer end , Dowty BJ, Carlisle DR and Laseter JJ. New Orleans Persons at High Risk of Cancer Ed. JF Fraumeni Jr.
seasing drinking water sources tested by gas chromatography- New York: Academic Press,Inc, p. 346,1975.
m*85 8Pectometry. Environstatal Science and 4 river'
~
7echnology 1975:9 762. Modan B, Barell V. Lubin F, Modan M, Greenberg RA p rectal 7 Environmental Protection Agency. Draft analytical and Graham S. Iow fiber intake as an etiologic p highly report: New Orleans area water supply study. factor in cancer of the colon. J Narl Cancer last 7, with Cincinnati, Ohio: Environmental Protection Agency,
- g 1975 55 15-18. .
1974. Axtell 134 and Chiane L. Jr. Changing relative frequency g from 8 Harris RH, Page T and Reiches NA. Carcinogenic of cancers of the colon and recturn in the United sy at Haurds of Organic Chemicals in Drinking Water. States. Cancer 1966;19,750-754.
24 g water Origins of Human Cancer Histt, Watson, Winsten. Wynder EL, Kajitani TS, Ishikawa S, Dodo H and Dncentr- Volume 4, Book A, pp 309-330 Cold Spring Takano A. Envirenmental factors of cancer of the l ttatt or Laboratory,1977. colon and rectum.11 Japanese epidemiological data.
at least W n TA and Nem JE. Relationships between Cantre 1969:23 1210-1220.
lmption alone. (Revised oersion received 24th October 1980)
Elimate ,
kuld be gest if
. . - LL 1
. p v
Lung Cancer .in Loulslana: Doath Certificato Analysis 3 Mrrisa S. Gotilleb, M.D., 3 ** 5 Linda W. Pickle, Ph.D., 5 Wililam J. Blot, Ph.D., 8 and Joseph F. Fraument, Jr., M.D. 6 7 02 E ,10 o 115 A: sin AcT-In a search for etiologic clues, a review was made residence at time of death, type of industry by which he af Cnth c ertificates of res! dents in a cluster of 1.oulalana or she was employed, usual occupation (kind of work Pirl:hes, mainly in the southern part of the state, where lung done during most of working life, even if retired),
c:nc t rnortality was high. A comparison of the statements on marital status, and parents' names. With the use of the
- 'upris n I r 3,327 patients with lung cancer and those of 3,327 1970 U.S. Bureau of the Census coding scheme (7), the cintrIls (matched by sex, race, age, and parish of residence) statements on occupation and industry were each (uring 1960-75 revealed an approximately twofold excess rist'. ..
assigned a three digit code. Each patient was then cecccirted with transportation equiprr ent manufacture, malnif classified inte 1 of 9 occupational and I of 22 m, dus-chipbullding, and the llahing industry. Smaller elevations of lung cinc r risk were found among older men who had been om. trial categories. We also examined the cerdficates to determine the likelihood of Acadian ancestry (probable, pl2y;d in petroleum exploration and production and among male and I: male residents of towns where the petroleum Industry was possible, or unlikely) by comparing birthplace and c m:Jor employer. In addition, Acadian ancestry was associated parents' names to compiled lists of names and updated with c higher risk of lung cancer among older male and female variants of original Acadian settlers recorded on ships' pl%nts.-JNCl 63: 1131-1137, 1979. ggg3 (g_y g, The usual residence listed on the certificate was used f r classification of each individual as to proximity to Exceptionally high mortality from lung cancer has e ch of seven primary types of manufacturing industry been reported among males in southern . buisiana in the area. The classifications were based on the parishes along the Gulf of Mexico, the Mississippi Predominant manufacturing industry in terms of years River, and the Atchafalaya Basin (1-3). Reasons for the ,
excess rates are not clear, but national correlation 3" **'51'"C' .and number of employees during 1940-65 (15) f r the individual's town of residence. In most of studies of lung cancer have suggested the role of occu- the smaller parishes, a town uniquely defined a up pational exposures, including shipbuilding, chemical, de a petroleum, and paper manufacturing industries (3, f) ',ne r ,rea, so that each subject was classified as located in buisiana. As a preliminary step to field was I cited in the up code area of residence and it was,
. studies in the high-risk parishes, a study was made of?i" 3 one m that areac -
occupational,nesidentialf and'e'llinfe#fanors-'gleliiied Odds rau.os were calculated (16,17) by sex and race from the death certificates of persons who resided in as esdmates of the relative nsk o!*1ung cancer associ-this area and died of lung cancer. ,
ated with each of the 22 andustrial categories,9 occupa-tional groups, residendal proximity to the 7 industries.
METHODS and 3 categories of Acadian ancestry, We examined Nineteen buisiana parishes (text-fig.1) were se- Positive associations in greater detail by calculating odds ratios for substrata and by using logistic models l lected for study because of elevated lung cancer death i rates among white males for the years 1950-69 and/or because of petroleum, chemical, or paper manufac-turing industries in the area. Of the 19 parishes,17 had AnsatvimoN Usm ICD =Internadonal Classificadon of Diseases.
mortality rates that exceeded the national average; in 10, the excess was over 30%. The parishes were typi-cally small, with 1970 populations ranging from 10,000 ' Received October 5,1978; revised April 6,1979; accepied May 8, to 145,000 (med,an, 30,000).
i 3979, A computerized listing of death certificates for 1960- Supported in part by Public licalth Service (PilS) contract N01-75 for the 19 parishes was provided by the Office of CP61058 (to M. S. G.) from the Division of Cancer Cause and
! Public Health Statistics, buisiana Health and Human Prevendon, Nadonal Cancer Institute (NCI).
- School of Medicine, Tulane University, New Orleans, La. 70112.
~
Resources Administration. Certificates for deaths at-
~
- louisiana IIcalth and lluman Resources Administration, New tributed to lung cancer [ ICD (5); 162.1,163 for 1960 Orle'n5- La. 70160.
l 67; ICD (6)- 162,1 for 1968-75] 'were identified and ' ddd"" "#'ines Dr. Gottlieb at the Department of Medicine, abstracted. Each certificate that reported lung cancer as the cause of death was matched by sex, race, age at [".'I'ne 7g ' University Medical Center,1430 Tulan death within 5 years, year of death within 1 year, and g,yi,,nmental Epidemiology Branch, NCI, National Institutes by parish of usual residence to a control death cerdfi- of 11ealth. PilS, U.S. Department of licalth, Eduation, and Welfare, j cate-i.e., one of a person who died from conditions Bethesda, Md. 20205.
' We thank Ms. Maureen Maguire, Department of Biostatistics, other than cancer. Johns flopkins University, for her computer program for the logistic Abstracted from each death certificate were the fol- analyses of ase-control data.
loiing: patient's name, birthplace, usual residence and 1131 JNci. vot_ 63. No. s. sortusta is7s
{
1 l
1132 G:till:b, Pickl , BI:t, cnd Frcum:nl ARK ANSAS r- - r- -
i
.,.-7 g i 3 ,3,/ ,, ,
! )e* h 8
Claisonst i UNeon
\ n u lQt e , r ** ~,
5h
'o e i -
/ LINCOLN .' .- M a k. . - OuaCHtra ,O i~ D I as, g s
\ s , -
! ,' g_ , y _ _
citavlttE J "5 -I .. /g (f-'
( $- ,* * m aoison ,
+
') REO ---*.L-
- ( 'g
- OL$010 %.,miv E R ', 'g i
l$ #* '7 i e
- g. '.- * #
weaa j.-.._f N ,<'>'+ *
-. ._ , {c c g
-g f
m 5,- t ,- . - . . i , LA i s' t'
~ ~ '
o Sasint !8, \
GRANT kg i 1
^
0 60 miles 4
l
$,, , l u, a.)3 o -
- ~~~.>
,- . 1. _ . g hi ?> u
>t r*l $ f j j vca=0a ,i " ' "
- E 5 j a,"voytttes i M l.S S l S S I P P 1 }{
La *
- r ..
I
- ?'"
e- ' %.s
- t s L
3,. e. i , ~T ; i
,,. J .e# 3 rgte, g p ,,, I .e 4W IN N l b ! ' Eva n. ', p 8
- CIANA 3 ," $
SEAustGan0 j L N eELint ,i sT. j 9," [',$7i n'~7- . . q x
k \.;-+ <._, t, t...y(t~r s r
.ou.c ; ! I ' '. 51
.,.. .\
, ,,,Gstg r g
, r..s. r,
- J JEFF. I a 4 -I \ , */7 '%. -
CA a U 1 ogygg gg' j S ,
_ e
] ,
_ ,,;._..e -
I
/ , .y \,.ua
._-}....---__.r. r . ., c s
+
g 3,
, Mo 4, *
, . ---m.- -
~_ .- r -c- -= .* --a ** y CAWER0N l leERlA7 - -' $; % ~ ~ ~ ' ~ ' '
vtausuon [i ,',,-).-.
,~ #g Q , 3 se .
e ,, e ri .', M.
s I , une -' b ', f.e L , Sy, rta e t
- . A o. *- M -
A"/q 7 l
TrxT.nceitt I.-Louisiana parishes included in the study (nomhauhed areas).
I to test for interaction and to obtain the maximum Indian males. The median age at death was 62.8 for likelihood point and interval estimates of the odds lung cancer padents and 61.4 for controls.
ratios after simultaneous adjustment for several factors Usual industry.-Information on 87% of the certifi- .:
l including age, marital status, year of death, birthplace, ates for white males, 60% for black males, 96% for j l and parish t>f residence [(18,19); Pickle LW, Rohde white females, and 93% for black females was sufficient l CA: Unpublished observations). Further details of the to classify the decedent's usual industry of emplay- f, statistical methods are in " Appendix-Statistical ment into 1 of the 22 industrial categories. Table 1 Methods." shows the number of male workers and the crude odds ratios for each industry. The odds ratios were signifi-RESULTS cmtly high for fishing and for transport equipment manufactere, primarily shipbuilding. Smaller elevated l In the 19 parishes during 1960-75, 3,327 people died ratios were associated with construction and transpor-of lung cancer: 2,162 white males,643 black males,372 tation service. Similar increases were seen for metal and white females, 146 black females, and 4 American machinery manufacture, oil refining, and business-JNO. Vol. 61. NO. S. NoVEbtBER 1979 l
l
i -
}
- Lung Crn%r in Lcuislina 1133 ,
l TA nt.E 1.-C.51dr ratios according to usual industry among males (both raers combined)
Odds ratio Industry Census code' Cases Controls (95% confi.
. dence limits)*
17-19, 27 459 459 1.00 (0.87, 1.16)
Agriculture -
40 1.81' (1.22, 2.69)
Fishing 28 72 49 128 116 1.11 (0.85, 1.44)
Crude oil extraction 47-48, 57-58 15 14 1.07 (0.51, 2.23)
Mining 303 1.16 (0.98, 1.37) 67-78 346 Construction 443 418 1.07 (0.92, 1.24)
M:nufacturing 0.71 (0.48, 1.04) 107-109 47 66 Lumber 26 1.27 (0.75, 2.14) j Mstal, machinery 147-207 33 219-238 29 13 2.20' (1.14, 4.25) i Transportation equipment 0.87 (0.65, 1.17) l 268-297 88 101 Food 1.05 (0.79, 1.40) 328-337 103 98 Peper 1.12 (0.70, 1.79) 347-369 38 34 Chemical 58 1.32 (0.93, 1.87)
Oil refining 377-378 76 119-138, 248-259, 299-319, 29 22 1.31 (0.75, 2.31)
Other 338-339, 337-398 407-429 201 161 1.27' (1.02, 1.58)
Trcnsportation service - 0.71 (0.46, 1.11)
Communications, utilities 447-479 35 49 507-598 49 59 0.83 (0.56, 1.22)
Whilesale ,
0.78' (0.63, 0.96) 607-698 165 209 Retail 0.67 (0.40, 1.11) 707-718 26 39 Finince 46 1.33 (0.90, 1.97)
Business-repair 728-759 61 Personal services, entertainment 769-809 52 45 1.16 (0.77, 1.74) 828-897 67 96 0.69' (0.50, 0.96)
Professional 0.76' (0.61, 0.95) 907-937 156 201 Public administration Unknown for: 274 287 0.95 (0.79, 1.14)
Whites 0.96 (0.76, 1.20) 256 263 Blacks 2,805 Total 2.805
- See (7).
- See " Appendix-Statistical Methods" for d2finition of interval estimates.
- Approximate 95% confidence interval does not include 1. ,
- s G- +-~' '-- - - -
j .--. . . ,y _
-.rw.; .-- - y - , _ _ .-: -: -
<- ~~-
TAa1.t 2.-Odds ration by race and age group for selected industries Race group Age group Odds ratio Odds ratio industry Co* AI'- Cases Co"' (95% confi.
Race
- Cases tmis (95% confi. yr tmls dence limits) dence limits)
W 65 38 1.7* (1.14, 2.60) <63 28 17 1.9* (1.00, 3.42)
Fishing 1.7* (1.02, 2.86)
B 7 2 3.0 (0.70, 13.08) 263 44 23 W 126 114 1.1 (0.85, 1.45) <63 75 87 1.0 (0.70, 1.34)
Cruda oil 1.0 (0.17, 6.00) 263 53 29 1.6* (1.03, 2.61)
B 2 2 W 286 256 1.1 (0.94, 1.36) <63 176 182 1.1 (0.88, 1.38)
Const uction 1.3 (1.00, 1.65)
B 60 47 1.3 (0.87, 1.95) 263 170 121 Metal, machinery W 31 25 1.2 (0.73, 2.12) <63 -20 20 1.1 (0.60, 2.12) !
B 2 1 1.7 (0.21, 13.21) 263 13 6 1.9 (0.71, 4.85) manufacture 2.7* (1.21, 6.22)
Tr nsportation-equipment W 25 13 1.9 (0.97, 3.73) <63 20 8 1.5 (0.53, 4.52)
B 4 0 == 263 9 5 me.nufacture 2.2 (0.94, 5.14)
Shipbuilding (code 228) W 21 12 1.7 (0.85, 3.53) <63 16 8 B 3 0 == 263 8 4 1.7 (0.53, 5.43)
W 35 31 1.1 (0.69, 1.85) <63 24 19 1.4 (0.77, 2.62)
ChImical mannfurture 1.0 (0.22, 4.55) .8 (0.40, 1.74)
B 3 3 263 14 15 Oil refining W 69 55 1.3 (0.88, 1.82) <63 38 38 1.1 (0.71, 1.79)
B 7 3 2.2 (0.59, 7.91) 263 38 20 1.7 (0.98, 2.94)
Transportation W 155 127 1.2 (0.97, 1.58) <63 103 89 1.3 (0.98, 1.79)
B 46 34 1.4 (0.87, 2.19) 263 98 72 1.2 (0.89, 1.69) i W 55 42 1.3 (0.87, 1.98) <63 33 28 1.3 (0.80, 2.23)
Business-repair 1.4 (0.76, 2.52)
B 6 4 1.4 (0.42, 4.97) 263 28 18 Auto repair (codes W 29 20 1.4 (0.81, 2.58) <63 20 11 ' 2.0 (0.96, 4.24)
- B 6 3 1.9 (0.49, 7.06) 263 15 12 1.1 (0.52, 2.38) 749-757)
- W - white; B = black.
- Approximate 95% confidence interval does not include 1.
JNO, VOL. 63. NO. 5. NOVEMBER 1979
1 34 G:till;b, Pickle, Cl::t, cnd Freum3ni , c repair. Decreased odds ratios were found for retail risk for chemical workers in white collar trades (7 sales and for professional and public administration cases, I controll was elevated, but a slighi deficit was seen for the remainder of the ' workers in this industry.
industries.
Table 2 lists odds ratios according to race and age The excess associated with fishing was slightly greater category for selected industries in which :he overall than twofold throughout most of the study area.
ratios were above 1. For nearly every industry shown, Although the total risk ratios for workers in whole-sale and retail sales were less than I, we analyzed death the risks were elevated in both whites and blacks. The certificates on which chemical, oil, or paper and/or ratios tended to be higher in blacks, though several in-dustries had few black workers. The major age differ- wood sales (as opposed to manufacture) weie men-ences were limited to the oil production and refining tioned. No increased risk was seen for chemical or pe-industries (excess rates were observed primarily for troleum sales workers. However, significant excess was people pver age 62) and to the transport and chemical associated with lumber wholesale and lumber and s manufacture and auto repair industries (excess rates furniture retail operations among whites, with 24 cases only at younger ages). and 10 controls [ odds ratio =2.35, approximate 95%
The logistic analyses suggested that odds ratios for confidence limits (1.12, 4.92)] reporting usual employ-certain segments of the study population were par- ment in this industry.
ticularly high. The ratio associated s th crude oil Occupation.-The odds ratios for major occupa-exploration and drilling rose to threefo among per- tional categories according to sex and race are given in i sons over age 62 in parishes with petro. am or paper table 3. Among males, the ratios were above I for f
industries. Although it was based on small numbers, the craftsmen and operators and below I for professional TAntz 3.-Odds ratios for usual occupation according to sez and race M ales Females ,
White Black White Black 3 Odds Odds Odds Odds Census ratio ratio ratio a ratio Occupation code- /
Con- (95% Con- (95% Con- (95% Con- (95%
Cases Cases Cases Cases trols confi-trols confi- trols confi- trols confi-dence dence dence dence limits) limits limits) limits) {
118 -- 0.58' -
-8 _ 16 - 0.51- t2.c 18 ~15 - 120 --- 4
- 1.68 - -%
Professional H)01-199 (0.43,'
~
(0.22, (0.60," ^
~
~~ " (0.217 "' ' i '
0.79) 1.19) 2.44) 13.42) 308 363 0.82* 6 19 0.33* 15 7 2.11 1 0 -
Managerial 201-299 (0.86, (0.70, (0.13 0.98) 0.81) 5.21) 301-399 50 65 0.77 2 1 1.67 14 4 . 3.31* 1 0 -
Clerical (1.11, (0.52, (0.21 1.12) 13.21) 9.85) 509 1.16* 43 30 1.46 1 0 - 0 0 -
Craftsmen 401-594 570 (1.01 (0.90, 2.37) l 1 34) '
203 1.27* 26 15 1,74 6 4 1.45 2 1 1.68 Operators 601-699 252 (0.21, (1.04, (0.91. (0.42, I
' 1.55) 3.33) 5.00) 13.42) 87 84 1.04 21 23 0.91 0, 0 -
0 0 -
Transporters 701-739 (0.50, (0.76, 1.42) 1.67) 1.10 329 316 1.08 2 1 1.67 6 3 1.90 Laborers 740-799 215 197 (0.49, (0.90 (0.87, (0.21 1.36) 1.36) 13.25) 7.30) 297 1.06 131 128 1.03 1 3 0.43 5 8 0.63 Farmers 801-899 312 (0.21, (0.89, (0.78, (0.06, 1.26) 1.36) 3.02) 1.94) 126 1.09 33. 33 1.00 17 25 0.67 40 30 1.45 Servicemen 901-989 137 (0.84, (0.85, (0.61, (0.35, 141) 1.65) 1.27) 2.52) 290 300 0 85 83 95 0.71 llousewives (0.59. (0.44, 1.22) 1.15) 200 0.79* 44 62' O.69 8 13 0.67 6 8 0.75 Unknown
- 161 (0.26, (0.63, (0.46, (0.27, 0.99) 1.04) 1.66) 2.20)
- See (7).
- Approximate 95% confidence interval does not include 1.
pCI Vol 63. No. s. NOVEMBER 1979
I Lung C:nc:r in L:ultns 1135 T.Aat.t 4 -Odds ration by ser and rare for residence in proximity to industry Males Females Black i White Black .
White Odds Odds Odds Industry Odds ratio ratio ratio ratio (census n- (95%
Con. (95% Con- (95% Con- (95% Cases code) Casa Cases confi- trols confi-trols confi- trois confi- trols dence dence dence -
dence limits) limits) limits) limits) 58 64 0.89 24 37 0.58 Stwmills (108-109) 384 393 0.97 168 157 1.10 (0.60, (0.33, (0.83, (0.85, .
1.42). 1.32) 1.05) 1.14) 0 3 3 1.00 0 -
Mital manufacture 25 15 1.65 0 0 -
(0.22, (087, (15&-168) 4.57) t 3.16) 7 1.73 +
30 23 23 1.00 12 Cenneries (278) 182 198 0 91 31 1.03 (0.55, (0.67, (0.74, (0.62, 1.74) 1.83) 4.49) t 1.13) 184 167 1.11 34 36 0.94 38 41 0.92 8 8 1.00 (
Grzin mills (279) (0.89, * (0.58, (0.57, (0.37, i 1.54) 1.48) 2.72) ,
1.39) 14 0.85 3't 35 0.94 12 [
P:per manufacture 204 216 0.94 42 47 0 89 (0.57, (0.38, (0.77, (0.57, (32&-329) 1.08) 1.56) 1.91) 1.15) 38 0.77 90 86 1.06 31 Chimical manufacture 441 485 0.89 134 148 0.88 (0.75, (0.44, (0.74, (0.67, (347-368) 1.15) 1.50) 1.33) 1.03) 28 1.88 58 46 1.31 10 6 1.66 Oil rsfineries (377) 306 270 1.16 15 (0.86, (0.60, (0.97 (0.99.
3.57) 2.00) 4.64) f 1.38) 0.97 69 74 0.92 49 36 1.54 Othtr 436 418 1.05 206 210 (0.63, (0.92, (0.91, .(0.77, 1.23) 1.23) 1.33) 2.58) (
372 372 146 146 Tctal 2.162 2,162 643 64'1 - ,
- r
- - n
~
-s-~--~ " -
p.,.-,e.
w.- . , - . ~ . - . ..
rnd managefiil' occupations. Amo'nk ' white' females. Proxirsity to industry.-Table 4 presents the number however, risk ratios higher than I were seen for of residents and odds ratios according to the nearness 4 of residence to 7 primary manufacturing industries.
professional, managerial, and clerical occupations. In Elevated risk esdmates for all 4 sex-race groups were f:ct, regardless of type of occupation, odds ratios for employed women were usually higher than for house-elevated only in towns where oil refineries were lo-cated. The odds ratios ranged from 1,16 [with approxi-wives.
I TAstz 5.-4tds ratios for Acadian ancestry according to sex and race Possible Acadian Non-Acadian Probable Acadian Odds Odds Odds i
ratio" -
ratio
- ratio
- Con- (95% Con- (95%
Subjects Con- (95% Casa Casa confi-confi- trols confi- trols trols dence dence dence limits) limits) limits) ,
443 442 1.02 1,562 1,591 0.93 157 129 1.24 White (0.97, (0.88 (0.82, males 1.19) 1.07) 1.59) 564 569 0.93 j
! Black 6 6 1.01 73 - 68 1.08 (0.76, (0.66, (0 33, ,
miles 1.55) 1.31) 3.08) 0.92 1.03 78 73 1.10 173 178 White 21 21 (0.65, (0.54, (0.75, f1 males 1.62) 1.32) ;
1.96) i 20 17 1.21 125 129 0.79 :
Black 1 0 -
(0.40, (0.60, j f: males 2.43) 1.57) l
- Values are relative to non-Acadian.
- V: lues are relative to possible or probable Acadian.
JNC3, Vol. 63. NO. s. NOVEMBEP.1979
1136' Gottilab, Picklz, Bist, etnd Frcumznl -
mate 95% confidence limits (0.97,1.38)] among white Petroleum production and refining are major indus-males to 1.88 (0.99, 3.57) among black males. The tries in Louisiana. and about 8% of the certificates for male residents listed these as the usual industry of the excess for males was not accounted for b) employment decedent. The oserall relative risk (1.2) was not large in the refineries, although the odds ratios were higher but rose to 1.7 for those at older ages. Further.
for those who also worked in the industry than for those who did not and was particularly evident in although the risk ratios were small and possibl3 influenced by factors that could not be accounted for in parishes that had both petroleum and chemical plants. this death certificate review, the higher ratios among Acadian ancestry.-A gradient of increasing risk as-sociated with the likelihood of Acadian ancestry was female as well as male residents of towns where refineries are located raise the possibility of community seen among white males (table 5). However, the in-was limited to patients of probable Acadian exposures to industrial carcinogens. Only limited data crease ancestry who died at age 65 or older [ odds ratio = 1.51, are available on the risks of cancer associated with the petroleum industry, but a recent study of mortality and approximate 95% confidence limits (1.07, 2.12)]. A morbidity statistics in Lm Angeles, where the industry similar pattern was observed for white female Acadians is a major employer, revealed a 60% excess risk of lung for whom a twofold risk [2.10 (0.65, 6.76)) was found cancer among petroleum and coal products wcakers among women over age 62. Although the original settlers were white, some black decedents were classified (23)
One of the stronger associations in this survey was as Acadian on the basis of criteria on birthplace and the excess lung cancer risk smong fishermen. An parents' names. He risk ratios were slightly above I approximate twofold increase among blacks and whites for blacks of possible or probable Acadian ancestry. was consistent with other reports (2f, 25), but an exPl anation of this finding is unclear. De higher risks DISCUSSION for auto repair, transportation services, and lumber and fumiture sales inctustries have been reported in other This survey of death certificates revealed a number of leads to lung cancer risk, but no single factor was surveys of lung cancer (23-25) and serve as clues to found to be responsible for the excess mortality rates possible occupational hazards, Acadiana encompasses an area of south central Loui-reported for this tumor in southern Louisiana. In a siana where Acadians settled in the 18th century and national correlation study, the high lung cracer mor- continue to live as a relatively isolated population; tality rates among males residing in corruies with pe- some still maintain an agrarian life-style unique to this troleum, chemical, paper, and transportation manufac-group. On the basis of family surnames on death certif-turing industries (3, f) suggested that occupational cicates, aboutz65-ni the white residents inathe entire <
-' factors ~ influenced.~ the geographic-patterns of - this
~
~
~
study area (or about '10% of ths~ whites 'in AEadiaria
tumor. The present study was conducted to determine whether the associations seen at the aggregate (county parishes) were classified as probable Acadians. About 20% of this group were farmers. Lung cancer risks were or parish) level also existed at the individual level and elevated approximately 1.5-fold among males and two-persisted in the 1970's. There was a twofold excess fold among females with probable Acadian ancestry among males in transportation manufacturing, mainly who died at 63 years of age or older. No excess lung shipbuilding; however, only slight increases were asso- cancer risk was observed at younger ages, perhaps a ciated with employment in the petroleum and chemical result of temporal changes in risk factors.
industries, and no increase was seen with paper und Although death certificate studies are a quick and pulp production. inexpensive means of assessing whether patterns seen An excess risk of lung cancer has been reported at the aggregate (parish or county) level reflect charac-recently among American men who were employed in teristics at the individual level, the data abstracted from shipyards during World War II (20). Although fewer the certificates are obviously limited. He statements than 1% of the certificates in this study listed ship- on industry and occupation are at best only a crude building as the usual industry, ship construction and description of the usual work, with no information as repair was one of the largest industries in Louisiana to detail, variety, or duratior of employment. Sim-during the war (21). A case-control study conducted in ilarly, residential histories are tnknown, and informa- j a high. risk area in coastal Georgia indicated that over tion on cigarette smoking (the inajor etiologic factor 20% of the patients recently diagnosed as having lung for lung cancer) is unavailable, so that causal in-cancer had worked in the shipyards that operated only ferences on specific risk factors cannot be made from >
during the war '20); shipbuilding was rarely listed as
- this study. Despite these limitations, death certificates the occupation of the decedent on death certificates represent a practical means of obtaining a large sample (22). If a similar situation is true in Louisiana, this extending back over a long term and may be a valuable ;
death certificate survey probably has underestimated hypothesis. generating resource. The clues obtained by the contribution of shipyard exposures to the area. wide this study may now be tested by case-control studies in lung cancer excess. Moreover, the parishes selected for high risk communities of Louisiana, where more de-this study did not include those in the immediate finitise information can be obtained from interviews i vicinity of New Orleans, where shipbuilding was a with lung cancer patients or their next of kin.
major industry, 1979 pct. vot 63. No. 5. Nottuarit .
8
' . i Lung C:nc;r in L:ulsl:na 1137 APPENDIX-STATISTICAL METHODS U.S counties with petroleum industries. Science 198:51-53.
1977 The odds ratios presented in the text are simple U) World Heahh Organizatio,t: International Classification of Dn.
cross. product ratios of the stratified 2X2 tables with an (7th rev). Geneva WHO.1965 (6) Pubhc Heahh Service, hational Cemer for Health Stausucs: . ,
adjustment of 0.5 added to each, count (16,17). Interval Intemauonal Danification of Diseases Adapted (8th rev).
I estirryates for these odds ratios,6, are of the form: exp DifEW Publ No.1693. Washington, D.C: U.S. Govt Print I
[in 612 (st of in 6)) with standard error defined by Off. 1968 g/(1/A+1/B+1/C+1/D), where A, B, C, and D are the (7) U.S. Bureau of the Census: 1970 Census of Population. Alpha.
adjusted counts for the 2X2 table. betical Index of Industries and Occupations. Washington.
The logisuc model used for further analysis included D.C: U.S. Govt Print Off. 1971 (s) AnstNact.T B: Histoire et Genealogie dn Acadiens. Quebec: Le parameters for ,various confounders and effect modi- Consiel de la Vic Frangaise en Amerique.1%5 fiers; this technique has been described in detail else- (9) UBLAsc D: The Acadian Miracle. Liayette, La.: Evangeline where [(19), equations 1-4). The inidal logisde models Pu Co I g : An Attempt to Re.assersble the Old Settlers contained mdicator parameters for age (<63 yr, 263 in Family Groups. Baion Rouge. La.: Claitor's Book Store, yr), marital status (marned, other), calendar year of 1965 death (1960-63,1964-67,1968-71,1972-75), birthplace ui) Ritota MP, GAcorT N: The Crew and Passenger Registradon (Louisiana, other state, foreign), parish of residence Lis.ts of the Seven Acadian Expeditions of 1785 (mimeograph grouping (by major industry), and interactions (effect "' 3
(,,, goj,][g, 3,'; 3,,;iianylarriages in France (Depanment of modifications) between these variables and the case- Ille.ca Vilaine, 1759-1776). Harvey, b.: New Orleans Bicen-control indicator. In addition, categorizations of the tennial Commission,1976 other variables of interest were included in the models UJ) Winausc OW: Acadian Odyssey. Baton Rouge, b.: Louisiana for each industry, e.g., terms representing the 3 Aca. State Univ Press 1955 dian ancestry Uf) Rosim AJ SR: Monial Seulen Along Bayou Mouche.
- . groups and three broad occupadon arte. 1770-1798, vol 11 (mimeograph series). Harvey, b.,1974 gories (white- and blue-collar workers and others). us) Louisiana State Depanment of Commerce and Industry: Direc.
Sequential deletions of terms were made to arrive at a iory of Manufacturers. Baton Rouge: b. State Dept Com.
final well. fitting model. Decisions concerning param. merce 8- Industry, 1942-75 eter deletion were based on the t-statistics for signifi- U6) W Ln B: On esdmating the relation between blood group and cance of the individual parameter estimates, on the d g7) a j"'["n " "' "
changes in the value of the log likelihood, and on the
""ir o an s ican e of the logarithm
-- of a ratio of frequencies. Ann Hum Genet 20:309-311, 1955 goodness of fit of the model as coinpared with the UB) Cox DR: Analysis of Binary Data. London: Methuen, 1970 observed stratified odds ratios. Interval esdmates for the U9) PamTict R:.Use of the logistic model in retrospective studies. ,
odds ratio are of the same form described above, where In d as the maximum likelihood estimateof-the-log of:" d("[.T Hn c .J t .Toupo A. .et. ah-Lung cancer ;- ~ -
- . after employment in World War Il shipyards. N Engl J Med the odds rano adjusted for any significant confounders, 299:620-624, 1978 with standard enors obtained from the inverse of the (21) Bureau of bbor Statistics: Wardme Employment. Produedon information matrix for each final model. These inter- and Conditions of Work in Shipyards. Dept Labor Bull No.
vals may be interpreted as either 2-unit support in- 824. Washington, D.C: U.S. Govt Print Off, 1945 tervals (26) or approximate 95% confidence intervals. (22) HAmntNciox JM, BLor WT, Hoorta RN, et at: Lung cancer in coastal Georgia: A death certificate analysis of occupation:
Bric/ communication. J Natl Cancer Inst 60:295-298, 1978 (23) MENcx HE, HtNDERsoN BE: Occupadonal differences in rates REFERENCES of lung cancer. J Occup Med 18:797-801, 1976 (2f) MtutAM S JR: Occupational Mortality in Washington State,
(/) mason TJ, McKAY FV: U.S. Cancer Mortality by County: 1950-1971. Nat! Inst Occup Safety Health DHEW Publ No'.
1950-69. DHEW Publ No. (NIH) 74 615. Washington, D.C: (NIOSH) 76175-A. Washington, D.C: U.S. Govt Pt.nt Off, U.S. Govt Print Off. 1974 1976 G) mason TJ, McKAv FW, Hoovan R. et at: Atlas of Cancer Mor. (25) The Registrar Gerieral's Decennial Supplement. England and tality in U.S. Counties: 1950-69. DHEW Publ No. (NIH) 75 Wales 1961 Occupational Monality Tables. London: HM Stat 780. Washington, D.C: U.S. Govt Print Off, 1975 Off. 1971 u) BLOT %T, FnAtmtN JFJn: Geographic patterns of lung can. (26) EDWARDS AW Jn: Likelihood: An Account of the Statistical cer: Industrial correlations. Am J Epidemiol 103:539-550, 1976 Concept of Likelihood and Its Application to Scientific Infer.
(f) BLOT hT, BRINToN LA, S:oNr BJ, et at: Cancer monality in ence. London, New York: Cambridge Univ Press,1972 JNCI. VOL 63. No. 5, NOVEMBER 1979