Document LKv3JdD8BMzn6YzwRQb1Myrnz

Lung Cancer in Louisiana: Death Certificate Analysis1 z S9 *1L+1I 01S Marise S. Gottlieb, M.D.. 13-4*4S* * Linda W. Pickle, Ph.D., 8 William J. Blot. PhD tnd Joseph F,, Fraumenl, Jr., M.O. *7 *8STRACT--In a *arch lor ellologtc duct, a review wa* made et death certificate* ol revident* in a Clutter of Louisiana pirtshss, mainly In the touthem part of the ttate, where lung cancer mortality wa* high. A comparison ol the (tatemenU on occupation lor 3.32? patient* with lung cancer and tho*e of 3,327 controls (matched by tea, race, age, and pariah of residence) during 1960-75 revealed an approximately twofold excess risk modeled with transportation equipment manufacture, mainly chipbuilding, and the fishing Industry. Smaller elevations of lung cancer risk were found among older men who had been em ployed hi petroleum exploration and production and among mate nd female residents of towns where the petrofeum Industry wts i mafor employer, (n addition, Acadian ancestry wa* associated efth e higher risk of king cancer among older male and female residents.--JNCI 63: T13t-1137. 1878. residence at time of death, type of industry by which he or she was employed, usual occupation (kind of work done during most of working life, even if retired), marital status, and parents' names. With the use of the 1970 U.S. Bureau of the Census coding scheme (7), the statements on occupation and industry were each assigned a three-digit code. Each patient was then classified into 1 of 9 occupational and 1 of 22 indus trial categories. We also examined the certificates to determine the likelihood of Acadian ancestry (probable, possible, or unlikely) by comparing birthplace and parents' names to compiled lists of names and updated variants ol original Acadian settlers recorded on ships' logs (-/*). The usual residence listed the certifies;., wz: Exceptionally high mortality from lung cancer has been reported among males in southern Louisiana parishes along the Gulf of Mexico, the Mississippi River, and the Atchafalava Basin (1-3). Reasons for the excess rates are not clear, but national correlation studies of lung cancer have suggested the role of occu pational exposures, including shipbuilding, chemical, petroleum, and paper manufacturing industries (3, 4) located in Louisiana. As a preliminary step to field 1. 'Studies in the high-risk parishes, a study was made of orcupational, residential, and ethnic factors gleaned from the death certificates of persons who resided in this area and died ol lung cancer. for classification of axh individual as to prcMtmm to each of primary types of manufacturing industry in tec area. The classifications were based on the predominant manufacturing industry in terms of years in existence and number of employees during 19-10-65 (75) for the individual's town of residence. In most of the smaller parishes, a town uniquely defined a zip code area, so that each subject was classified as living "near" a particular type of industry only if a factors was located in the zip code area of residence and it was the major one in that area. Odds ratios were calculated (16, 17) by sex and race as estimates of the relative risk ol lung cancer associ ated with each of the 22 industrial categories, 9 occupa METHODS tional groups, residential proximity to the 7 industries, and 3 categories of Acadian ancestry. We examined I. Vineteen Louisiana parishes {text-fig. 1) were se lected for study because of elevated lung cancer death positive associations in greater detail by calculating odds ratios for substrata and by using logistic models rates among white males lor the years 1950-69 and/or because of petroleum, chemical, or paper manufac turing industries in the area. Of the 19 parishes, 17 had ) monalitv rates that exceeded the national average; in 10, the excess was over 30V The parishes were typi ( cally small, with 1970 populations ranging from 10,000 f to 1 -15.000 (median, 30.000). A computerized listing of death certificates for 1960- ; 75 for the 19 parishes was provided by the Office of Public Health Statistics, Louisiana Health and Human i Resources Administration. Certificates for deaths at tributed to lung cancer (JCD (5); 162.1, 163 for 1960- 67; ICD (6): 162.1 lor 1968-75] were identified and abstracted. Each certificate that reported lung cancer as the cause of death was matched by sex, race, age at j death within 5 years, year of death within 1 year, and I bv parish of usual residence to a control death certifi cate--i.e., one of a person who died from conditions f other than cancer. i Abstracted from each death certificate were the fol lowing; patient's name, birthplace, usual residence and ! Abbri.'' iaTios tsin ICI> = International Classification of Diseases. 1 Received October 5. revised April 6. 1979; accrpied Mas 8. 1979. } Supported in pan b> Public Healin Service (PHS) comrsc. NOi. CP61058 (to M. S G.) Irom (he Division M Cancer Cause and Prevention, National Cancer Institute (NCI). * School of Medicine, Tulane University, New Orleans, La. * Louisiana Health and Human Resources Administration. New Orleans, La. 70160 * Address reprtnts to Dr. Gottlieb ai the Department of Medicine. Tulane University Medical Center, HS0 Tulane Ave., New Orleans, La. 70112. 4 Environmental Epidemiology Branch. NCI, National Institutes ol Health, PHS. U.S. Department of Health, Education, and Welfare, Bethesda. Md. 20205. ? We thank Ms. Maureen Maguire, Department of Biostatisiics, Johns Hopkins University, for her computer program for the fogisiic analyses of case-control data. 1131 J\a. VOL. 63. NO. 5, NOVEMBER 1979 *i \ 1132 Gottlieb, Pickle, Blot, and Fraumenl ARKANSAS T \\' I Tt\TF!Ci RE I.--Louisiana parishes included in the smdv (cross-hatched areas). to test for interaction and to obtain the maximum likelihood point and interval estimates of the odds ratios after simultaneous adjustment for several factors including age, marital status, year of death, birthplace, and parish of residence [f/fi. 19)\ Pickle L\V, Rohde CA: Unpublished observations]. Further details of the statistical methods are in "Appendix--Statistical Methods." RESULTS In the 19 parishes during 1960-75, 3,327 people died of lung cancer: 2,162 white males, 643 black males. 372 white females. 146 black females, and 4 American Indian males. The median age at death was 62.8 for lung cancer patients and 61.4 for controls. Usual industry.--Information on 87% of the certifi cates for white males, 60% for black males, 96% for white females, and 93% for black females was sufficient to classify the decedent's usual industry of employ ment into 1 of the 22 industrial categories. Table 1 shows the number of male workers and the crude odds ratios (or each industry. The odds ratios were signifi cantly high for fishing and for transport equipment manufacture, primarily shipbuilding. Smaller elevated ratios were associated with construction and transpor tation service. Similar increases were seen for metal and machinery manufacture, oil refining, and business- JNCl. VOL. M. NO S. NOVEMBER 1979 Lung Cancer In Louisiana 1133 TaBLE 1 --Odds ratios according to usual industry among moles (troth races combined) Industry Apriculture Fishing Cruce oil extraction Mining Construction Manufacturing Lumber Metal, machinery Transportation equipment Food Paper Chemical Oil refining Other Transportation service Communications, utilities Wno.esale Retail Finance Business-repair Personal services, entertainment Professional Public administration Lnxnown for. V." bites Blacks Total Census code* 17-19. 27 28 49 47-48. 57-58 67-78 107-109 147-207 219-238 268-297 328-337 347-369 119-138, 248-259, 299-319. 338-339. 337-398 407-429 447-479 507-598 607-698 707-718 728-759 769-809 828-897 907-937 Cases 459 72 128 15 346 443 47 33 29 88 103 38 76 29 201 35 49 165 26 61 52 67 156 274 256 2.805 Controls 459 40 116 14 303 418 66 26 13 101 98 34 58 22 161 49 59 209 39 46 45 96 201 287 263 2.805 Odds ratio (95% confi dence limits)' 1.00 <0.87, 1.16) 1.81' (1.22. 2.69) 1.11 (0.85. 1.4 41 1.07 (0.51. 2.23) 1.16 (0 98. 1.37) 1.07 <0.92. 1.24) 0.71 <0.48. 1.04) 1.27 (0.75. 2.14) 2.20' (1.14. 4 25) 0.87 (0 65. 1.17) 1.05 (0.79, 1.40) 1.12 <0.70. 1.79) 1.32 (0.93. 1.87) 1.31 (0.75. 2.31) 1.27' (1 02. 1.58) 0.71 (0.46. Ml) 0.83 (0.56. 1.22) 0.78' (0.63. 0 96) 0.67 (0.40. l.lll 1.33 (0.90. 1.97) 1.16 (0.77. 1 74) 0.69' (0.50. 0.96) 0.76' (0.61. 0.95) 0.95 (0.79. 1.14) 0.96 (0.76. 1.20) * Set (D. ' Sci 'Appendix--Statistical Methods' for definition of interval estimates, ' Approximate 95% confidence interval docs not include 1 10S Table 2 -- Odds ratios by race and age group jar selected industries Industry Race* Race group Cases Con trols Odds ratio (95% confi dence limits) Age. yr Age group Cases Con trols Odds ratio (95% confi dence limits) FL-.-.;:.g Crude oil Construction Metal, machinery manufacture Transportation-equipment manufacture Shipbuilding (code 228) Chemical manufacture Oil refining ` 4 Transportation Business-repair Auto repair (codes 749-757) \V B W B W B W B \Y B W B W B W B W B W B W B 65 38 1 7' (1.14, 2.60) <63 28 17 1.9* (100. 3.421 7 2 30 (0.70, 13 08) >63 44 23 1.7' (1.02. 2.86) 126 114 1 1 (0.85. 1.45) <63 75 87 1 0 (0.70. 1.34) 2 2 1.0 (0 17, 6.00) >63 53 29 16* (1.03. 2.61) 286 256 1.1 (0.94. 1 36) <63 176 182 1.1 (0.88. 1.38) 60 47 1.3 (0.87. 1.95) >63 170 125 13 <1.00. 1 65) 31 25 1.2 (0.73. 2 12) <63 20 20 1.1 (0.60, 2.12) 2 1 1.7 <0.21. 13.21) >63 13 6 1.9 (0.71, 4.85) 25 13 1 9 (0 97, 3.73) <63 20 4 0 OC >63 9 8 2.7* (1.21, 6.22) 5 1.5 <0.53. 4.52) 21 12 1.7 (0.85. 3.53) <63 16 3 0 SC >63 8 8 2.2 (0.94. 5.14) 4 1.7 (0.53. 5.43) 35 31 1.1 (0.69. 1.85) <63 24 19 1.4 (0.77. 2.62) 3 3 1 0 (0.22. 4.55) >63 14 15 .8 (0.40. 1 74) 69 55 1.3 (0.88. 1.82| <63 38 38 1.1 (0.71, 1.79) 7 3 22 (0.59. 7.91) >63 38 20 1.7 (0.98. 2.94) 155 127 1.2 (0.97, 1.58) <63 103 89 1.3 (0.98. 1.79) 46 34 1.4 (0.87. 2.19) >63 98 72 1.2 (0.89, 1.69) 55 42 1.3 (0.87, 1.98) <63 33 28 1.3 (0.80. 2.23) 6 4 1.4 <0.42. 4.97) >63 28 18 1.4 (0.76. 2.52) 29 20 1.4 (0.81. 2.58) <63 20 11 2.0 (0.96, 4.24) 6 3 1.9 (0.49. 7.06) >63 15 12 1.1 (0.62, 2.38) *W = white; B" black. * Approximate 95% confidence interval does not include 1. JNCI. VOL. 63. NO. 5. NOVEMBER 1979 34 Gottlieb, Pickle, Blot, and Fraumenl repair. Decreased odds ratios were found for retail sales and for professional and public administration industries. Table 2 lists odds ratios according to race and age category for selected industries in which the overall ratios were above 1. For nearly every industry shown, the risks were elevated in both whites and blacks. The ratios tended to be higher in blacks, though several in dustries had few black workers. The major age differ ences were limited to the oil production and refining industries (excess rates were observed primarily for people pver age 62) and to the transport and chemical manufacture and auto repair industries (excess rates only at younger ages). The logistic analyses suggested that odds ratios for certain segments of the study population were par ticularly high. The ratio associated with crude oil exploration and drilling rose to threefold among per sons over age 62 in parishes with petroleum or paper industries. Although it was based on small numbers, the risk for chemical workers in white-collar trades rr, cases, I control) was elevated, but a slight deficit wj, seen for the remainder of the workers in this industry The excess associated with fishing was slightly greats than twofold throughout most of the study area. Although the total risk ratios for workers in whole, sale and retail sales were less than 1, we analyzed death certificates on which chemical, oil, or paper and/or wood sales (as opposed to manufactuie) were men. tioned. No increased risk was seen for chemical or pe. troleum sales workers. However, significant excess was associated with lumber wholesale and lumber and furniture retail operations among whites, with 24 cases and 10 controls [odds ratio = 2.35, approximate 955 confidence limits (1.12, 4.92)] reporting usual employ mem in this industry. Occupation.--The odds ratios for major occupa tional categories according to sex and race are given in table 3. Among males, the ratios were above 1 fot craftsmen and operators and below 1 for professional Mt . i Ci- STO 11*7468 OccuD&tion Professional Managerial Clerical Craftsmen Operators Transporters Laborers Farmers Servicemen Housewives Unknown Census code * 001-199 201-299 301-399 401-594 601-699 701-739 740-799 801-899 901-989 Table 3.--Odds ratios for usual occupation according to ter and race Cases Males White Con trols Odds ratio (95% confi dence limits) Cases Black Con trols Odds ratio (95% confi dence limits Cases Females White Con trols Odds ratio (95% confi dence limits) Cases 70 118 0.58* 8 16 0.51 18 15 1.20 2 (0.43. (0 22. (0.60. 0.79) 1.19) 2.44) 308 363 0 82* 6 19 0.33* 15 7 2.11 1 (0.70, (0.13. (0.86. 0.98) 0.81) 5.21) 50 65 0.77 2 1 1.67 14 4 3.31* 1 (0.52, (0.21. (1.11. 1 12) 13.21) 9.85) 570 509 1.16* 43 30 1.46 10 0 (1.01. (0.90. 1.341 2.37) 252 203 1.27* 26 15 1.74 6 4 1.45 2 (1.04, <0.91. (0 42. 1.55) 3.33) 5.00) 87 84 1 04 21 23 0.91 00 0 (0.76, (0.50. 1.42) 1.67) 215 197 1 10 329 316 1.08 (0.90. (0.87, 2 1 1 67 <0.21, 6 1.36) 1.36) 13.25) 312 297 1.06 131 128 1.03 1 3 0.43 5 (0.89. <0.78. <0.06. 1.26) 137 126 1.09 1.36) 33 33 1.00 3.02) 17 25 0.67 40 (0.85. (0.61. <0.35. 1 41) 1.65) 1 27) "" 290 300 0.85 83 (0.59. 1.22) 161 200 0 79* 44 62 0.69 (0.63. (0.46. 8 13 0.67 6 (0.27. 0.99) 1.04) 1.66) Black Con trols Odds ratio <95% confi dence limits) i 1.68 (0.21. 13.42) 0 0 0 1 1.68 (0.21. 13.42) 0 3 1 90 (0.49. 7.30) 8 0.63 <0.21. 1.94) 30 1.45 (0.84. 2.52) 95 0.71 (0 44. 1.15) 8 0.75 (0.26. 2.20) * See (T). * Approximate 95% confidence interval does not include 1. JNCI. VOL. 63. NO. S. NOVEMBER 1979 31!. he pr< far: err. w v t V I Lung Cancer (n Louisiana 1135 TaBIE 4.--Odds ratios by ter and race for residence in proximity to industry Males White Industry (census code) Cases Con trols Odds ratio (95% confi dence limits) Cases 7*-mills (108-109) teul manufacture Zmnenes <278) ^rain mills (279i Paper manufacture (32S-329) Chemical manufacture ` (347-368) Oil refineries (377) Other Total 384 25 182 184 204 441 306 436 2.162 393 15 198 167 216 485 270 418 2.162 0.97 (0.83. 1.14) 1.65 <0.87. 3.16) 0.91 (0.74. 1.13) 1.11 (0.89. 1.39) 0.94 (0.77. 1.15) 0.89 (0.74, 1.03) 1 16 (0.97. 1 38) 1.05 (0.91, 1 23) 168 0 31 34 42 134 28 206 643 Black Con trols Odds ratio (95% confi dence limits) 157 1.10 (0.85, 1.42) 0 30 1.03 (0.62. 1.74) 36 0.94 <0.58, 1.54) 47 0.89 (0.57. 1.38) 148 0.88 (0.67, 1.15) 15 1.88 (0.99, 3.57) 210 0.97 (0.77. 1.23) 643 Cases 58 3 23 38 33 90 58 69 372 Females White Con trols Odds ratio (95% confi dence limits) Cases 64 3 23 41 35 86 46 74 372 0.89 <0.60. 1.32) 1.00 (0.22. 4 57) 1.00 (0.55. 1.83) 0.92 <0.57. 1.48) 0.94 (0.57. 1.56) 1.06 (0.75. 1.50) 1.31 (0.86. 2.00) 0.92 (0.63, 1.33) 24 0 12 3 12 31 10 49 146 Black Con trols Odds ratio (95% confi dence limits) 37 0.58 (0.33, 1.05) 0 7 1.73 (0.67. 4.49) 8 1.00 (0.37. 2.72) 14 0.85 (0.38. 1.91) 38 0.77 (0.44, 1.33) 6 1.66 (0.60. 4.64) 36 1.54 (0.92, 2.58) 146 STOI 47469 jnd managerial occupations. Among while females, however, risk ratios higher than 1 were seen for professional, managerial, and clerical occupations. In fact. regardless of tvpe of occupation, odds ratios for Employed women were usuallv higher than for house wives Proximity to industry.--Table 4 presents the number of residents and odds ratios according to ihe nearness of residence to 7 primary manufacturing industries. Elevated risk estimates for all 4 sex-race groups were elevated oniy in towns where oil refineries were lo cated. Tlie odds ratios ranged from 1.16 [with approxi- i ' Table 5--Odds ratios for Acadian ancestry according to sex and race 1 . Subjects White males Black males White ' females Black females _ Probable Acadian Cases Con trols Odds ratio* (95% confi dence limits) 157 129 1.24 (0.97. 1.59) 6 6 1.01 (0.33. 3.08) 21 21 1.03 (0 54. 1.96) 1 0-- Possible Acadian Cases Con trols Odds ratio* (95% confi dence limits) 443 442 1.02 (0.88. 1.19) 73 68 1.08 (0.76. 1.55) 78 73 1.10 (0.75. 1.62) 20 17 1.21 (0 60, 243) Cases 1,562 564 173 125 Non-Acadian Con trols Odds ratio* (95% confi dence limits) 1.591 569 178 129 0.93 (0.82. 1.07) 0.93 (0.66. 1.31) 0.92 (0.65. 1.32) 0.79 (0.40. 1.57) ' Values are relative to non-Acadian. Values are relative to possible or probable Acadian. JNCI. VOl.. 63. NO. 5. NOVEMBER 1979 1136 Gottlieb, Pickle, Blot, and Fraumenl mate 95% confidence limits (0.97, 1.38)] among white males to I 8S (0.99. 3.37) among black males. The excess lor males teas not accounted lor by employment in the refineries, although the odds ratios were higher for those who also worked in the industry than for those who did not and was particularly evident in parishes that had both petroleum and chemical plants. Acadian ancestry.--A gradient ol increasing risk as sociated with the likelihood of Acadian ancestry was seen among white males (table 5). However, the in crease was limited to patients of probable Acadian ancestry who died at age 63 or older [odds ratio= 1.51, approximate 95% confidence limits (1.07, 2.12)]. A similar pattern was observed lor white female Acadtans for whom a twofold risk (2.10 (0.65, 6.76/] was found among women over age 62. Although the original settlers were white, some black decedents were classified as Acadian on the basis of criteria on birthplace and parents names. The risk ratios were slightly above 1 for blacks of possible or probable Acadian ancestry. DISCUSSION This survey of death certificates revealed a number of leads to lung cancer risk, but no single factor was found to be responsible for the excess mortality rates reported for this tumor in southern Louisiana, fn a national correlation study, the high lung cancer mor tality rates among males residing in counties with pe troleum. chemical, paper, and transportation manufac turing industries (3, 4) suggested that occupational factors influenced the geographic patients of this tumor. The present study was conducted to determine whether the associations seen at the aggregate (countv or parish) level also existed at the individual level and persisted in the 1970's. There was a twofold excess among males in transportation manufacturing, mainly shipbuilding; however, only slight increases were asso ciated with employment in the petroleum and chemical industries, and no increase was seen with paper and pulp production. An excess risk of lung cancer has been reported recently among American men who were employed in shipyards during World War 11 (20). Although fewer than \% of the certificates in this study listed ship building as the usual industry, ship construction and repair was one of the largest industries in Louisiana during the war (21). A case-control study conducted in a high-risk area in coastal Georgia indicated that over 20% of the patients recently diagnosed as having lung cancer had worked in the shipyards that operated only during the war (20); shipbuilding was rarely listed as the occupation of the decedent on death certificates (22). If a similar situation is true in Louisiana, this death certificate survey probably has underestimated the contribution of shipyard exposures to the area-wide lung cancer excess. Moreover, the parishes selected for this study did not include those in the immediate vicinity of New Orleans, where shipbuilding was a major industry. % Petroleum production and icfining an major inttu,. ArF"x:: tries in Louisiana, and about 8% of the certificates fw The male residents listed these as the usual industry of the ci OSS r. decedent. The overall relative risk (1.2l was not largf adjourn but rose to 1.7 for those at older ages. Further, although the risk ratios were small and possibly influenced by factors that could not be accounted for in this death certificate review, the higher ratios among fit 11 Mao. t? v< 92 1. A5 female as well as male residents of towns where 1 he refineries are located raise the possibility of community para me exposures to industrial carcinogens. Only limited data fiers; tl are available on the risks of cancer associated with the where ( petroleum industry, but a recent study of mortality and coiuaii morbidity statistics in Los Angeles, where the industry >t;. nr is a major employer, revealed a 60% excess tisk ol lung j death cancer among petroleum and coal products workers | (Liuis (23). gr< >i| One of the stronger associations in this survey was the excess lung cancer risk among fishermen. An CO approximate twofold increase among blacks and whites was consistent with other reports (24, 25), but an CD explanation of this finding is unclear. The higher risks for auto repair, transportation services, and lumber and furniture sales industries have been reported in other surveys of lung cancer (23-25) and serve as clues to possible occupational hazards. Acadiana encompasses an area of south central Loui CD siana where Acadians settled in the 18th century and continue to live as a relatively isolated population; some still maintain an agrarian life-style unique to this ob- gToup. On the basis of family surnames on death certif od. icates, about 6% of the white residents in the entire hi 5 study area (or about 10% of the whites in Acadiana the - parishes) were classified as probable Acadians. About w: - ' 20% of this group were farmers. Lung cancer risks were elevated approximately 1.5-fold among males and two in: fold among females with probable Acadian ancestry [ (t: v who died at 63 years of age or older. No excess lung cancer risk was observed at younger ages, perhaps a result of temporal changes in risk factors. RtT Although death certificate studies are a quick and inexpensive means of assessing whether patterns seen I < at the aggregate (parish or county) level reflect charac teristics at the individual level, the data abstracted from the certificates are obviously limited. The statements on industry and occupation are at best only a crude description of the usual work, with no information as to detail, variety, or duration of employment. Sim 1ilarly. residential histories are unknown, and informa tion on cigarette smoking (the major etiologic factor ifor lung cancer) is unavailable, so that causal in ferences on specific risk factors cannot be made from this study. Despite these limitations, death certificates i represent a practical means of obtaining a large sample (extending back over a long term and may be a valuable (hypothesis-generating resource. The clues obtained by this study may now be tested by case-control studies in high-risk communities of Louisiana, where more de i finitive information can be obtained from interviews ( with lung cancer patients or their next of kin. ; JSCI. VOL. 65. SO. S. NOVEMBER 1979 11 *11 1o 1 s Lung Cancer In Louisiana 1137 APPENDIX--STATISTICAL METHODS The odds ratios presented in the text are simple no-produci ratios o( the stratified 2X2 tables with an jdjusunent of 0.5 added to each count (16, 17). Interval (stimates for these odds ratios, , are of the form: exp r/n 82 (se of In )] with standard error defined by v'( i ,4+T'B+T C+l/D). where A, B, C, and D are the adjusted counts for the 2x2 table. The logistic model used for further analysis included parameters for various confounders and effect modi fiers; this technique has been described in detail else where [(19), equations 1-4]. The initial logistic models contained indicator parameters for age (<63 yr, >63 vr). marital status (married, other), calendar year of death (1960-63, 1964-67, 1968-71, 1972-75), birthplace (Louisiana, other state, foreign), parish of residence grouping (by major industrv), and interactions (effect modifications) between these variables and the casecontrol indicator. In addition, categorizations of the other variables of interest were included in the models for each industry, e.g., terms representing the 3 Aca dian ancestry groups and three broad occupation cate gories (whue- and blue-collar workers and others). Sequential deletions of terms weie made to arrive at a final well-fitting model. Decisions concerning param eter deletion were based on the t-statistics for signifi cance of the individual parameter estimates, on the changes in the value of the log likelihood, and on the goodness of fit of the model as compared with the observed stratified odds ratios. Interval estimates for the odds ratio are of the same form described above, where In is the maximum likelihood estimate of the log of the odds ratio adjusted for anv significant confounders, with standard errors obtained from the inverse of the information matrix for each final model. These imer1 vals may be interpreted as either 2-unit support in tervals (26) or approximate 95 confidence intervals. REFERENCES (/) MstON TJ. Mf.Kw FW: L'.S. Cancer Mortality bv Countv: 1950-69. DHEW Publ No. (MH) 71-615. Washington. D C..: L'.S. Govt Print Off, 1974 Q) Mason TJ. McKay FW. Hoover R, et al: Allas erf Cancer Mor tality in L'.S. Counties: 1950-69. DHEW Publ No. (NTH) 75780. Washington, D C: L'.S. Govt Print Off, 1975 0) Blot WT. Fraimem JF Jr: Geographic patterns of lung can cer: Industrial correlations. Am J Epidemiol 105.539-550. 1976 (4) Bi.OT WT. Biunton LA. Stone BJ. et al: Cancer mortalits in L'.S. counties with petroleum industries. Science 198:51*53. 1977 (5) World Health Organization: Iniernational Classification of Dis eases (7th rev}. Geneva: WHO. 1965 (6) Public Health Service, National Center for Health Statistics: International Classification of Diseases Adapted (8th tew. DHEW Publ No. 1693. Washington. D.C.: U.S. Govt Prim OfE 1968 (7) L'.S. Bureau of the Census: 1970 Census of Population Alpha betical Index of Industries and Occupations. Washington. DC.: U S. Govt Prim Off, 1971 () Arsenault B: Histoire et Genealogie des Acadiens. Quebec Le Consiel de la Vie Frangaise en Amerique, 1965 (9) LeBlanc D: The Acadian Miracle. Lafayette. La.: Evangeline Publ Co.. 1966 (10) March and SA Sr: An Attempt to Reassemble the Old Settlers in Kamils Groups. Baton Rouge. I^a.r Claiior's Book Store, 1965 {11) Rjkdlr MP. Galdet N: The Crew and Passenger Registration Lists of the Seven Acadian Expeditions of 1785 (mimeograph series}. Metairie, La.. 1965 (12) Robichaux AJ Sr: Acadian Marriages m France (Department of Ille-et Vilaine, 1759-1776). Harvey. La.; New Orleans Bicen tennial Commission, 1976 (J2) WiZERtiNC OW: Acadian Odvssey. Baton Rouge, La.: Louisiana State Univ Press. 1955 (14) RobichaL-x AJ SR: Colonial Settlers Along Bayou Lafouche. 1770-1798. voi II (mimeograph senes). Harvey. La.. 1974 (15) Louisiana State Department of Commerce and Industry: Direc tory of Manufacturers. Baton Rouge: La. State Dept Com merce fc Indusiry, 1942-75 06) Woolfe B. On estimating ihe relation between blood group and disease Ann Hum Genet 19.251-253. 1954 (17) Haldane JB The estimation and significance of the logarithm of a ratio of frequencies. Ann Hum Genet 20:309-311, 1955 OS) Gox DR Analysis of Rinan Data. London: Methuen, 1970 (79) Prentice R- Use of the logistic model in retrospective studies. Biometrics 32.599-606. 1976 (20) Blot WT. Harrington* JM. Toledo A, et al: Lung cancer after employment in World War II shipyards. N Engl J Med 299 620-624, 1978 (21) Bureau of Labor Statistics: Wartime Employment. Production and Conditions of Work in Shipyards. Dept Labor Bull No. 824 Washington. DC: U.S. Govt Print Off. 1945 (22) Harrington JM. Blot WT. Hoover RN. et al: Lung cancer in coastal Georgia: A death certificate analysis of occupation Brief communication. J Natl Cancer Inst 60:295-298. 1978 (23t Menck HE. Henderson BE: Occupational differences in rates of lung cancer. J Occup Med 18:797-801. 1976 (241 Mii.mvm S Jr: Occupational Mortality in Washington State. 1950-1971. Nail Inst Occup Safety Health. DHEW Publ No. (NIOSH) 76-175-A. Washington, D.G: U.S. Govt Print Oil, 1976 (25) The Registrar General's Decennial Supplement. England and Wales 1961 Occupational Mortality Tables. London: HM Star Off. 1971 (26) Edwards AW JR: Likelihood: An Account of the Statistical Concept of Likelihood and Its Application to Scientific Infer ence. London. New York: Cambridge L'mv Press. 1972 JNCI. VOL 63. NO. 5. NOVEMBER 1979