Document VGQDwmaNRnpJvkBL6kRxbX3BK

A Follow-up Study of a Ten-Percent Sample of the Canadian Force. 1. Cancer Mortality in Males, 1965-731.2 Geoffrey R. Howe and Joan P. Lindsay4. 4BSTRACT-A system is described for monitoring the mortality ex- perience of an approximately 10% sample of the Canadian Labor Force Nith the use of computerized record linkage to the Canadian mortality lata base. Results are presented for the 4.203 cancer deaths that xcurred between 1965 and 1973 in 415,201 males with known xcupational histories from 1965 to 1969. A number of previously suggested associations between certain occupationsand cancers have ieen confirmed in these results, and a number of new associations ~ l s ohave been observed. The system provides a powerful tool for both tenerating and testing hypotheses, and this power will increase as wther mortality experience is accumulated by the cohort.-JNCI 1983; 70:37-44. Many associations between certain occupations and in:reased risk of various cancers have been reported ( / ) . -:stirnates of the percentage of cancer attributable to OCCUiational exposure have ranged as high as 40% ( 2 ) ,though his figure is a matter of some dispute ( 3 ) . Most of the nown occupation-cancer associations initially developed -om anecdotal observations, and little work has been done :I the area of systematic monitoring of large samples of the ibor force to detect new possible associations. ThP- are two major problems in attempting to establish llch nonitoring system: I ) the cost of assembling a dlkiZ6tly large cohort of individuals with known occupalonal history and 2) the dificulty and cost in following uch a cohort for a suficient period of time to obtain data n a large enough number of cancer cases or deaths. T h e ;osest approach to scch a system is that of the U.K. Legistrar General who issues a decennial supplement ( 4 ) . mxiations between occupation and cancer or other causes f death are estimated as follows: The number of deaths :ported by occupation as recorded on the death certificate used as the numerator, and the estimated number of ldividuals working in that occupation during the time in (Jestionis used as the denominator. This method, of course, LS a drawback. Because of the time lag involved, the >serveddeaths are not necessarily occurring in the working Wlation that constitutes the denominator. This paper reports the establishment of a monitoring .'stem based on individuals (forming an =IO% sample of le Canadian labor force) for whom occupational histories etwen 1965 and 1971 are available. Computerized record n b g e ( 5 ) is being used in conjunction with the Canadian lortdity data base maintained by Statistics Canada (for- .%' Dominion Bureau of Statistics) to determine the iQrtahtyexperience of the cohort. Initial findings of the "dY are reported for cancer mortality occurring in male 'l e m b e ~of the cohort between 1965 and 1973. The plan is continue mortality monitoring in the future, and the should thus provide a very powerful tool for both and testing hypotheses concerning associations .'twelrbccupation and cancer risk. MATERIALS AND METHODS Occupational histories.-Between 1965 and 1968, the Canadian UIC collected data from all employers in Canada for all employees whose SI number ended in the digit 4. T h e data consisied of the individual's SI number, surname, sex, year of birth, and occupation and industry in which the individual was currently employed. For 1969and 1971,only those persons whose SI number terminated in the digit 4 preceded by an odd-numbered digit were included in the sample. The data for 1970 were not available for the purposes of the present study. The occupations were coded with the use of the three-digit 1961 Canadian census occupation codes (6) for the data from 1965 to 1969 and the four-digit 1971 Canadian census codes ( 7 ) for the 1971 occupations. Industries were coded with the use of corresponding versions of the 1961 and 1971 three-digit Canadian census industry codes (8, 9). The data were converted to machine-readable form and were originally used to examine employment patterns in Canada (IO) and thus originally were not intended for use in epidemiologic research. The file of approximately 3 million records was sorted by SI number to bring together records referring to the same individual, thereby forming composite occupational histories. Records that had the same SI number but differed in surname, sex, or year of birth were inspected, and as a consequence less than 0.1% of the records were suficiently discrepant to indicate that they referred to different individuals. The discrepant records were dropped from the file, leaving 700,335 individuals. To obtain further identifying information for the purposes of record linkage, a list of the SI numbers was given to the Canadian C'IC, the agency responsible for the issuing of SI numbers. The UIC added identifving data for all members of the cohort, so that the items available for linkage were surname: first and second given names; sex; day, month, and year of birth: mother's AEEREVlATtOXS USED: ICD=lntcrnational Chsification of Diseases: ICD7=7th revision of ICD: ICD-8~8threvision of ICD: Sl=social insurance; SMRIstandard mortality ratio(s):UIC=Uncmplovment Insurance 6 m mission. ' Received March 16. 1982; accepted August 4. 1982. *Supported by a grant from the National Cancer Institute of Canada. National Cancer Institute of Canada Epidemiology Unit, Facultv of Medicine, University of Toronto, Toronto. Ontario. Canada, .M5S 1.48. 'Health Division. Statistics Canada, Ottawa. Ontario. Canada. K 1.4 025. 'We arc indebted to Dr. H. B. Newcombe, formerly of Atomic Energy ofCanada Ltd., on whose original suggestion the present study was based. We also thank Mr. R. Thibault, the Unemployment Insurance Commission of Canada. for his assistance in providing identifving data from the social insurance number mater index file. 37 JNCI, VOL. 70, NO. 1, JANUARY 1983 t 38 Howe and Lindsay maiden name; and most recent province in which the individual worked. Mortaliy fillow-up.-Statistics Canada maintains a mortality data base that contains records for all deaths registered in Canada since 1950 ( / I ) .These data are supplied to Statistics Canada by the vital statistics registrars in the various provinces and territories. The data also include the deaths of Canadian residents that occurred in the United States, which probably account for most of the extraterritorial deaths, To determine the mortality of the occupational cohort, records from the cohort were linked to the mortality data base for the years 1965-73 (the latest years for which data were available at the time the linkage was performed). .Methods for the computerized linkase of records on the basis of such identifying characteristics as surname and date of birth originally were developed in Canada by Newcombe (12) and his associates. These methods were developed further by Howe and Lindsay (5) to form a generalized iterative record linkage system. which is now being used routinely by Statistics Canada for a number of medical follow-up studies (13). Record linkage involves the comparison of all records in one file (the occupational cohort) with all records in another file (the mortality data base) and the calculation of a score (weight) that is a measure of the relative probability that two records, one from each file, refer to the same person. The weights are calculated by comparison of the individual identifying items, such as given names, surname, or day, month, and year of birth, and assignment of a positive or negative component to the weight depending on whether the items in question agree, partially agree, or disagree. The components of the weight are based on the relative frequency of the value of the identifying item and on an estimate of how frequently the item in question is misrecorded. The methods and computer system that were used have been described in detail (5).When a match occurred between a member of the cohort and a death record, the date and cause of death [coded to the appropriate revision of the ICD (14, 1.5))were abstracted from the death record and added to the occupational information for that individual. The data for analysis, therefore, consisted of records for 415,201 males who entered the cohort between 1965 and 1969. Those who entered in 1971 (42,135) were excluded because the 1961 and 1971 census occupation codes are not readily compatible. Thus 1971 occupations were not considered in the present analysis. Females also were excluded because of the relatively small number of deaths that occurred during the period in question. A more meaningful analvsis of the data for females will be possible when further years of mortality data are available. Statistical analysis.-SMR were computed for various subgroups of the cohort (as defined below) for all causes of death combined, for all cancer deaths, and for deaths ac- 'All data processing involving individual records containing identifying data was done in Statistics Canada under the auspices of the Canadian Federal Statistics Act. .Analyseswere performed after the identifying information had been removed from the records. thus preserving the privacy of individuals in the rcudv. 4cording to 19 cancer categories (table 1). Expected num -.of deaths were calculated by application of the a g e - s p a rate for the particular cause of death for the reference to the person-years of observation of the exposed gro 15 age categories (0-19, 20-24, . . . 80-84,r85).The for a particular cause of death and for a particular e group was then calculated in the usual way as the the observed number of deaths in the group to the e number. It was not necessary to control for c observation inasmuch as mortality rates wer the time period under study. .A member of t defined as entering the study (Le., contributing person of observation) at the midpoint of the year in which pational information was first recorded for him a exiting the study either in the year of death or at the en 1973 if still alive. S M R were computed with respect both to Cana population death rates ( / 6 ) and to death rates for occupational cohort as a whole. Because Canadian popug tion death rates reflect the expected "healthy worker" e ( / 7 )(see below), and because it is preferable when rn data are ascertained by record linkage that both of interest and the referent group havr ieir determined by record linkage (14, the S,MI. pres generally those with respect to the entire cohort significance and interval estimation were p treatment of the observed number of deaths variable and with the use of standard techniques the Poisson distribution. In interpreting the statistica nificance of the results presented, however, the prob confounding the multiple comparisons must be tak account, as discussed subsequently. RESULTS Table 1 shows (in terms of ICD codes) the various causes of death considered and the number of de cribed to those causes that occurred in the cohort. gives results of comparison of the mortalitv expene the entire cohort with that expected from comparison Canadian population rates. T h e mortality rate for the c from all causes combined is approximately 20% lower that occurring in the general population, the size o reduction being very similar to that reported in other occupational cohorts and presumably accounte healthy worker effect. The death rates from all for lung and bladder cancers (two sites that hav of postulated occupational associations) are no the same extent from the population rates. This indicates a larger (age-adjusted) proportion of dea lung and bladder diseases in the cohort as cornpa number of deaths from these causes in the genera tion-again a not unexpected result. To do asystematic examination of the data, i t is nec of course, to define the "exposure" groups. A n options are available since for each individual in t potentially 10 variables are available that define t pational history, i.e., the occupation and industry co up to 5 years. For this study, an individual was cate as exposed to a particular occupation or industry if been recorded as being employed in that occupa JNCI, VOL. 70, NO. I , JANUARY 1983 TABLE 1.-Number of cancer deaths, by site, in cohort, 1-73 Cancer site ac. -. bronchus, and lung mach estine except rectum wreas ner neoplasms of lymphatic ind hematopoietic tissue >stategland idder plus other urinary 'rgans' :turn and rectosigmoid unction Jkemia and aleukemia ccal cavity and pharynx ,xcept lip !in ,phagus er I'.W n ne -at ier and unspecitied sites Total ICD-7 code" 162-163 151 152-153 157 200-205 177 181 154 204 141-148 193.0 150 155 161 190-191 196 140 170 140-205 ICD-8 codeb 162 151 152-153 157 200-203 NO. Of deaths 1,336 408 359 253 245 185 188-189 207 194 154 162 204-207 141-149 154 138 191-192.0 150 155-156 161 172-173 1i o 140 174 140-209 128 87 81 56 52 30 7 3 303 4.203 ICD-7 code applies to deaths that occurred from 1965 to 1968. ICD-8 code applies to deaths that occurred from 1969 to 1973. In subsequent tables, bladder cancers (ICD-7 code 181.0 and ICDfide 188) and other urinary organ cancem are listed both separately cornbined. ustry for at least 1 year between 1965 and 1969; thus he :t cluded in more than one exposure group. Analysis j N -,,lade of the data for individuals who to be listed as osed had to have 2, 3, 4, or 5 years of recorded employI t in the particular occupation or industry. Analyses of records for these individuals inevitably yielded smaller .Ibers for the various exposure groups, but the results tined were essentially similar to those reported here. lough in many cases a degree of overlap exists between iin occupations and industries, for the present analyses ipation and industry were examined independently. able 3 shows the SMR for the 13 major occupational sions into which the 1961 census codes are grouped (6), 111 causes of death combined, for all cancers, and for the .ading individual cancer causes of death. Bladder cancer other urinary tract cancers are listed both separately combined. Table 4 gives the corresponding data for the najor industry divisions. Tables 5 and 6 show results .ing to individual three-digit occupation and industry 's. An association was included when the S M R was ated. with at least five observed deaths and the nominal lficance level was 50.05 (one-sided). Inevitably, limitations become apparent in performing the type of monitoring exercise described in this paper: 1) We had to assume that among individuals employed in a particular occupation or industry there would be those who had been in that occupation or industry, or a similar one, for a number of years. This assumption was necessary primarily because of the latent period effect, since the present data related to mortality occurring within at most 8 years of the first recorded exposure to a particular occupation or industry. Also, a minimum number of years of exposure may be necessary for a person to be at increased risk. although numerous studies indicate that for many occupation-cancer associations the minimum period of exposure necessary is very small ( / ) . Inclusion in an exposure group of individuals who had not achieved any necessary latent period or minimum time of exposure would, of coume, lead to a misclassification bias. Such bias would lead to an underestimation of risk, and the estimate thus would be conservative. 2) Lack of data concerning possible confounders as smoking or socioeconomic factors makes it impossible to distinguish indirect from direct associations. 3) With 274 different occupation codes, 274 industry codes. and 19 different cancer causes of death, a problem with multiple comparisons inevitably arises, in the sense that at least 1 in 20 of the associations recorded (tables 5 , 6) may be due to chance. Because of the limited number of years of death records available for study and the low mortality of the cohort as a whole, several established occupation-cancer associations are not yet apparent for this cohort. However, some significantly high SMR correspond with findings in other occupational studies: High mortality due to cancer of the brain in occupation division 2 (professional and technical occupations: table 3) corresponds with the findings of the 1971 U.K.occupational mortality decennial supplement ( 4 ) . In occupation division 5 (service and recreation occupations), an elevated SLUR for caccer of the buccal ca-;ity and pharynx except lip is similar to the U.K. cancer registrations data, which show high rates of tongue and mouth cancer ( 4 ) . Industry division 10 (community, business, and personal service industries: table 4) also has a high rate of cancer of the buccal cavity and pharynx except lip. reflecting the fact that many of the same individuals had jobs classified in occupation division 5 (table 3). In addition. two separate occupations in division 5 , bartenders and waiters (table 5), show elevated SMR for buccal cavity and pharynx except lip. Both the data presented here and the V.K. data ( 4 ) show high mortality from all cancers in transport and communi- TABLE 2.-SMR for various causes of death in males in cohort, 1965-73 CUSSION ) u r primary purpose in preparing the present paper was :port on the establishment of the monitoring system. refore, extensive results and comparisons with results 1 other similar monitoring systems are not presented . ir, yuch as such data will be more meaningful when 'r I. .bers of deaths are available for analysis. Cause of death All causes All cancers Lung cancer Bladder cancer No. of deaths Observed 19,374 4,315 1,370 124 Expected 23,399 4,881 1.443 139 90% c o d SMR" dence interval 0.83 0.82-0.84 0.68 0.864.90 0.95 0.91-0.99 0.89 0.761.a "Computed with reference to Canadian population rates, 1965-73. JNCI, VOL. 70, NO. 1, J A X U A R Y 1983 40 Howe and Lindsay II 1 1. JNCI, VOL. 70, NO. 1, JANUARY 1983 Lanaaian LaDor torce Study 1 41 JNCI, VOL. 70, NO. 1, JANUARY 1983 - 42 Howe and Lindsay TABLE 5.-SMR and number of observed deaths for males, by three-digit occupation code Bartenders Bricklayers, stonemasons, and tile setters Butchers and meat cutters Canvassers and other door-to-door salesmen Carpenters Civil engineers Clerical occupations, NES Commerical travelers COOkS Driver-salesmen Farm laborew Fishermen Fitters and assemblers (metal) Hoistmen, cranemen, and derrickmen Inspectors, examiners. and gaugers (metal). NES Linemen and servicemen (telephone, telegraph, and power) Lumbermen, including logging laborers Mechanical engineers Mechanics and repairmen, (motor ve- hicles) Messengers Metal drawers and extruders Metalworking machine operators. NES Metalworking occupations. NES Owners and managers. NES Painters (construction and maintenance). paperhangers, and glaziers Plumbers and pipe fitters Sales clerks Service station attendants Stationary enginemen Telegraph operatom Truck drivers Waiters Warehousemen and freighthandlers, NES Buccal cavity and pharynx except lip Trachea, bronchus, and lung AU malignant neoplasms Trachea, bronchus, and lung Rectum and rectosigmoid junction Other neoplasms of lymphatic and hematopoietic tissue AU malignant neoplasms Other neoplasms of lymphatic and hematopoietic tissue AU malignant neoplasms Buccal cavity and pharynx. except lip Intestine except rectum Stomach Intestine except rectum Leukemia and aleukemia AU malignant neoplasms Stomach Trachea, bronchus, and lung All malignant neoplasms Other neoplasms of lymphatic and hematopoietic tissue Trachea, bronchus. and lung Pancreas Trachea, bronchus, and lung Intestine except rectum Stomach All malignant neoplasms Trachea. bronchus, and lung .U malignant neoplasms Esophagus Bladder Bladder plus other urinary organs AU malignant neoplasms Trachea, bronchus, and lung AU malignant neoplasms Leukemia and aleukemia Stomach Urinary organs other than bladder Bladder plus other urinary organs AU malignant neoplasms Trachea, bronchus, and lung Buccal cavity and pharynx except tip Trachea, bronchus, and lung Prostate gland Stomach Bladder plus other urinary organs Bladder plus other urinary organs All malignant neoplasms Leukemia and aleukemia Buccal cavity and pharynx except lip .v1 malignant neoplasms Prostate gland "NES=not elsewhere specfied. bComputed with reference to entire cohort. ' = 0 . 0 1 ~ . 0 5 ," = 0 . 0 0 1 ~ . 0 1 , '**=P=s0.001. "No. of observed deaths, 3 5 . Nominal -0.05 (one-sided). 6.35." 2.13'' 1.75"' 1.76. 4.40' 2.72. 0.85. 0.41' 2.48' ' 1.79' 1.39' 0.67. 1.91' 3.24' 1.32" 2.61" 0.55' 0.65'' 3.22' 1.70' 2.60' 1.62' 3.53' ' 2.33' 0.74" 0.59' 2.39' 3.59" 2.36' 2.17' 1.82.'. 1.82' 3.90' 2.47' 2.02' 2.21' 1.67' 0.85' 0.71** 2.85' 1.68' 2.58". 0.41. 5.29**' 2.45' 1.99' 1.69. 7.18". 1.41'' 2.35' 6 19 48 12 5 5 189 5 10 I8 39 21 12 6 85 9 10 36 5 13 5 16 8 6 67 17 8 7 6 9 39 I2 7 7 8 8 16 I79 4a 5 24 17 5 7 7 9 13 12 74 10 cation occupations (division 6), although this is not the case for industry division 7 (transporation, communication, and other utilities). Within this occupation division, commercial travelers suffered high mortality rates from cancer of the intestine except rectum, as did commercial travelers United Kingdom (I).Both sets of data showed mortality due to cancer of the trachea, bronchus, for messengers. The postulation has been made JNCI, VOL. 70, NO. 1, JANUARY 1983 'I Canadian Labor Force Study 1 43 TABLE 6.-SMRand number of observed deaths for males, by three-digif indusby code - Industry E products &..,age manufacturers Bowling alleys and billiard parlors Building construction Piah products Fishing Food stores General merchandise stores Grain elevators Hqhway and bridge maintenance Hospitals Hotels, restaurants,and taverns Household furniture Insurance and real estate agencies Insurance carriers Iron and steel mills Legal service Lowe Machine s h o p Manufacturers of major appliances (electricand nonelectric) Men's Ciothing Metal stamping, pressing, and coat- ing Motor vehicle dealers -%io. ehicle manufacturers Motor vehicle repair shops Other crop and livestock combina- tion farms h t h g and publishing W a y transport &tail stores. NESd Savings and credit institutions services incidental to agriculture special trade contractors Tadcab operations Tmck transport veneer and plywood mills m o h a l e r s of food Cancer site Intestine except rectum Buccal cavity and pharynx except lip All malignant neoplasms Trachea: bronchus. and lung All malignant neoplasms Other neoplasms of lymphatic and hematopoietic tissue Intestine except rectum All malignant neoplasms All malignant neoplasms Trachea, bronchus, and lung Rectum and rectosigmoid junction Trachea, bronchus, and lung All malignant neoplasms Brain M malignant neoplasms Buccal cavity and pharynx except lip Larynx Brain Intestine except rectum Intestine except rectum All malignant neoplasms All malignant neoplasms All malignant neoplasms Trachea, bronchus, and lung Trachea. bronchus, and lung Other neoplasms of lymphatic and hematopoietic tissue All malignant neoplasms Stomach Intestine except rectum All malignant neoplasms Esophagus Other neoplasms of lymphatic and hematopoietic tissue Leukemia and aleukemia All malignant neoplasms Stomach .411 malignant neoplasms Intestice except rectum Urinary organs other than bladder Bladder plus other urinary organs 0th neoplasms of lymphatic and hematopoietic tissue Bladder plus other urinary organs Bladder plus other urinary organs Stomach AU malignant neoplasms Stomach .411 malignant neoplasms Lar?.nx All malignant neoplasms All malignant neoplasms Prostate gland "Computedwith reference to entire cohort. f'=o0.01rI;W.05, "..0.001~0.01, ***=Ps0.001. .No. of observed deaths, r5. Nominal -0.05 (one-sided). Nm=not elsewhere specified. SMR" 2.63" 5.501.40' 3.01'. 0.78". 0.40'. 0.32". 0.62' 0.71' 0.56. 2.97' 0.51" 0.32" 2.94 1.70' 3.44*** 2.66' 4.61" 2.69' 3.66" 1.28' 2.16' 0.76'' 0.62" 1.85' 5.28" 0.51'. 2.27' 2.25' 1.30' 2.75' 2.31" 2.82' 1.45 2.75' 0.54 3.22' 2.68" 1.78' 0.38' 7.75. ' 2.59' 3.75" 1.51' 0.48' 1.56' 4.83" 1.20' 0.45" 2.54' No.of observed deaths' 10 5 37 7 203 6 7 19 39 10 6 14 5 5 16 20 7 5 5 6 71 8 95 25 12 5 11 8 7 48 5 13 6 35 5 10 9 11 19 5 5 6 6 25 10 28 5 104 8 6 rate for messengers in the United Kingdom may be '**selection by individuals of this kind of work ( 4 ) . ,A rate of stomach cancer among farmers and farm "orken (occupation division 7) was also evident for the malain industry division 1 (agriculture) and for those listed as laborers in table 5-clearly due to the overlapping Of' t' categories. ,A similar observation was reported in studies in Japan and Iceland (19, 20) but not in the U.K. ( 4 ) decennial supplement. No significantly elevated mortality levels were found at the divisional level for craftsmen, production processors, and related workers. However, within the division, bricklayers, stonemasons, and tile setters, as well as plumbers and pipe fitters, showed elevated mortality due to trachea, bronchus, JNCI, VOL. 70, NO. I , JANUARY 1983 & Howe and Lindsay and lung cancer. Similar elevations also were apparent in the U.K.data ( 4 ) , but the high levels disappeared when controls for social class were instituted, indicating that lifestyle rather than occupational hazards may be responsible for the elevated mortality. In addition to these observations, which are supported by other studies, other potentially interesting associations in the present results (tables 5 , 6) seem worthy of further investigation. The three limitations discussed above imply that the appropriate uses of a monitoring system such as the system described here are twofold: 1) The data may be used to generate hypotheses that then can be tested further in appropriate analytical case-control or cohort studies and 2) the data may be used to test hypotheses generated from other studies. In tests of hypotheses suggested in other studies, it will be possible, of course, to define an exposure group much more specifically in terms of an appropriate combination of occupation and industry codes and years of exposure. Although a number of interesting results have emerged from the present analysis, the power of the monitoring system to detect associations will increase substantially as further years of death data are accumulated. For this reason, the plan is to continue monitoring the mortality experience of the cohort on a regular basis. REFERENCES ( I ) MirLER GL. Indexes to xlcctcd literature on occupational and environmental carcinogenic hazards. Philadelphia: The Franklin Institute, 1975. ( 2 ) BRlDBORD K. DECOUFLE P, FRAUMEJNFI. et al. Estimates of the fraction of cancer in the United States related to occupational facton. Bethada. .Md.: National Cancer Institute, National Institute of Environmental Health Sciences. and National Institute for Occupational Safety and Health, 1978. (3) HIGCINSOJN. Present and future developments in environmental carcinogenesis. In: .Advances in medical oncology, research and education. Vol 111. Oxford: Pergamon Press, 1979:203-113. ( 4 ) Registrar General for England and Wales. Occupational m o r t + , 1970-72: Decennial supplement. London: Her Majatv's Stat 1978. (5)HOWE CR. LixosAy J. A generalized iterative record linkage m- puter system for use in medical follow-up studies. Comput B i o w R a 1981; 14:327-340. (6) Dominion Bureau of Statistics. Occupational classification m a w census of Canada. 1961. Ottawa: Dominion Bureau of Stat- -1961. (7) . Occupational classification manual census of Canada, 1971. -Vol I. Ottawa: Dominion Bureau of Statistics. 1971. ( 8 ) Standard industrial classification manual. 1960. Ottay, Dominion Bureau of Statistics. 1960. ( 9 ) -. Standard industrial classification manual, 1970. Otta- Dominion Bureau of Statistics, 1970. (10) NEWCOMBHEB. A method of monitoring nationally for possibk delayed effects of various occupational environments. he Committee on Scientific Criteria for Environmental Quality, &. p01-t No. NRCC 13686. Ottawa: National Research Council Canada, 1974. (11) SMITH ME, NEWCOMBHEB. Automated follow-up facilities in Can& for monitoring delayed health effects. Am J Public Health 1%: 70: 1261- 1268. ( 1 2 ) NEWCOMBHEB. The use of medical record linkage for population and genetic studies. Methods Inf Mcd 1969: 8:7-1 I. 13) SMITHME, SILINSJ, LINDSAY JP. The present state of automata follow-up in Canada. 11. The products. J Clin Comput 1980; 9:1% 30. 1 4 ) World Health Organization. International classification of diseas, Geneva: WHO, 1957. 1x7,15) -. International classification of disease. Geneva: M'HO. 16) Statistics Canada. Causa of death. Ottawa: Statistics Canada, 1 s 1973. 17) M C ~ ~ I C H AJE. LStandardized mortality ratios and ,he "heal&! w19o7rk6e;r18e:f1fe6c5t-"1:68S.cratching beneath the surface. J Occupat h4cd ! 18) HOWE GR. LINDSAY J, COPPOCKE, MILLER AB. Isoniazid e x p u n & relation to cancer incidence and mortality in a cohon oftubertub patients. Int J Epidcmiol 1979: 8:305-312. I( 19) TSUCHIKC.ARole of environmental factors in cancer pr0ductim.J Kcio Med Soc 1967: 44:709-717. (20) SIcuRjo"ssoN J. Chronological study of stomach cancer. bladid 1969: 55:Il7-127. JNCI, VOL. 70, NO. 1, JANUARY 1983 7'-1. -... -2. Y 9-