GASTRIC CANCER AND OCCUPATION:  A REVIEW OF THE LITERATURE

BY: SUSAN R STOCK MD MSc FRCPC
FOR: Industrial Diseases Standards Panel (Occupational Disease Panel)
N.B.: REVISED SEPTEMBER 1994


TABLE OF CONTENTS

1
PREFACE

2
INTRODUCTION

3
THEORIES OF OCCUPATIONAL STOMACH CARCINOGENESIS

4
REVIEW OF THE LITERATURE
         4.1 METHODS
         4.2 EPIDEMIOLOGIC CONSIDERATIONS
         4.3 STUDIES REVIEWED

5

IS THERE EVIDENCE OF A RELATIONSHIP BETWEEN STOMACH CANCER
AND OCCUPATION?
6 EVIDENCE FOR OCCUPATIONALLY RELATED STOMACH CANCER IN
CANADIAN GOLD MINERS
7 IS THERE EVIDENCE OF A CAUSAL RELATIONSHIP BETWEEN STOMACH
CANCER AND OCCUPATIONAL EXPOSURE IN ONTARIO GOLD
MINERS?
8 CONCLUSIONS

9
RECOMMENDATIONS
10 APPENDIX A     ELERAVANT REFERENCES IDENTIFIED BUT NOT
REVIEWED

1 PREFACE

This review of the occupational health literature was commissioned by the Ontario Industrial Diseases Standards Panel to addressing the following questions:

1) is there evidence of a relationship between stomach cancer and occupation?
2) is there evidence of a causal relationship between stomach cancer and any specific occupational exposures?

Although it was commissioned in order to assist the panel in its deliberations on stomach cancer in gold miners, the author was asked not to focus on gold or other miners but rather look at all occupations in which stomach cancer risk may be elevated to address the very broad question of whether there is evidence that any workplace exposures can contribute to the development of gastric cancer. It was not possible to negotiate a more limited question. The author was asked to examine the hypothesis of dust as a possible etiologic agent. A forty hours consulting time limit was imposed by the IDSP.

In order to address the above questions it is important that an unbiased and as comprehensive as possible identification of relevant studies be carried out and reviewed. Unfortunately this was not possible and the current review is much more limited in scope and not exhaustive. An initial bibliographic search on Medline identified 86 relevant English language articles. The IDSP staff were aware that it was not possible to read, critically appraise, and write a comprehensive review of the full international literature on occupational stomach cancer within a 40 hour limit and specifically stated that they did not want a meta-analysis or similar qualitative systematic review of the topic. This review is therefore an initial overview of the topic. It is not an exhaustive and systematic overview in which all potentially relevant papers are identified, reviewed, and systematically evaluated for methodologic quality, and conclusions are drawn based on the highest quality studies. Rather it aims to provide members of the Panel a general overview of some of the salient studies, some of the most significant methodologic issues that threaten the validity of the conclusions of these studies, a brief overview of theories of the biological basis of occupational stomach carcinogenesis, and provide tentative conclusions based on a limited review.

Ontario gold miners are believed to have been exposed to silica, arsenic, chromium, diesel emissions, mineral fibre (actinolite, tremolite), radon decay products and aluminum powder1. Based on the set of potential exposures of gold miners, the review targeted stomach cancer studies in miners, stomach cancer studies of workers potentially exposed to high levels of dusts, particularly those containing silica or metals such as arsenic or chromium. There is substantial literature linking stomach cancer and asbestos and/or mineral fibre and the topic is a huge one in and of itself and was felt to be beyond the scope of this review. The topic of carcinogenesis associated with ethylene oxide is vast and unlikely to be relevant to gold miners. Studies addressing stomach cancer and exposures such as ethylene oxide or asbestos were therefore not included in this review.

A large number of additional relevant studies were identified in the course of reviewing the bibliographies of the papers reviewed. Because of the time constraints these papers were not reviewed but are identified in Appendix A.

2 INTRODUCTION

Gastric cancer was the most frequent cancer in the world for both genders in 19802 and continues to be one of the most common cancers worldwide3. It is the most common cancer in Japan and China. In North America and Europe, although its incidence has been declining considerably over the past 50 years, it is still a relatively common cancer. In the US the 1982-1986 average annual age adjusted incidence rate for males was 12.3 and for females 5.5 and mortality rate was 8.4 for males and 3.8 for females. The National Cancer Institute of Canada and Statistics Canada have estimate 3000 Canadians will develop stomach cancer in 1994, two-thirds among males and that 2,110 Canadians will die of stomach cancer this year. Stomach cancers make up 3.9% of all cancer deaths among Canadian males (the sixth most common site of cancer deaths) and 2.9% of female cancer deaths. The 1994 age-standardized mortality rate for stomach cancer among Canadian and Ontario males is estimated at 6/100,000 and among females at 3/100,000 (similar to leukaemia). Mortality rates for stomach cancer have been decreasing at an average annual rate of 3.3% between 1982 and 1991. Although incidence is declining it is not an insignificant problem.

To date few specific causes of stomach cancer have been clearly identified. Demographic trends and changes in stomach cancer rates following migration have provided strong evidence that have been clearly identified. Dietary hypotheses include high salt intake, pickled foods, nitrate intake, a lack of fresh vegetables, and food cooked on coal burning cookers likely to be high in PAHs. Infection with helicobactyr pylori has been shown to cause acute gastric inflammation, increase the risk of chronic gastritis, atrophy, peptic ulcer and gastric cancer. Stomach cancer in the general population tends to occur in older individuals and is mostly a disease of those over 60 or 70. It is consistently associated with lower socioeconomic status but no specific determinant associated with low socioeconomic status has been identified. It has been found to be higher among a wide range of immigrants in North American or European populations but none of these studies simultaneously controlled for socioeconomic status. Most frequently it has been associated with East or Central European origin/ethnic identity. In some studies relevant to Canada rates have been found to be elevated in French Canadians and in Newfoundlanders. Family studies have found a 2-4 fold risk increase of stomach cancer in relatives of cases suggesting a potential genetic component. It is highly likely that stomach cancer has multifactorial causation and that environmental factors are of considerable importance. The most common environmental exposures leading to stomach cancer likely come from nutritional sources. Occupational exposures are also highly plausible as potential causes because a number of carcinogens can be ingested in high quantities following inhalation or skin contact of the hands (hand to mouth route).

3 THEORIES OF OCCUPATIONAL STOMACH CARCINOGENESIS

Although there are probably at least 100 studies which have examined a link between occupation and stomach cancer, few articles in the literature have attempted to explain potential mechanisms for occupational gastric carcinogenesis. The main theories found in the literature have arisen in an effort to explain stomach cancer in coal miners and were described by Ames who suggested the following hypotheses:

1. Poly-aromatic hydrocarbons (PAHs) especially benzopyrene may be adsorbed by coal mine dust, adsorbed precarcinogens may be activated by the p-450 oxidase enzyme system and lead to intragastric production of carcinogens that attack mucosal tissue and induce gastric carcinoma. PAHs such as benzopyrene are known to transform epithelial cells and initiate cancer. Combustion products of coal have been shown experimentally to be carcinogenic to the gastric epithelium in mice and in tissue culture (Weinberger et al). A similar PAH hypothesis is proposed by Nutt4 as a likely explanation of the stomach cancer found in rubber workers. PAHs are also found in lubricating oils and are a possible explanation of increased stomach cancers found in bearing plant workers found by Park et al and Silverstein et al. Interestingly the sharp decline in gastric cancer in many developed countries parallels the drop in the use of coal burning in homes and this cancer is high in those countries where coal burning for cooking is still common.

2. Exposure to dust in general may cause gastric cancer. Respirable dust may be subject to macrophage scavenging, brought up through mucociliary clearance and swallowed as with other inhaled particles. Coal mining dust exposures include iron dust, iron oxides from welding, other metals found in coal, rock dust, particulate from blasting, asbestos insulation and coal shuttle brake linings dust, diesel particulate. Coal miners may also have been exposed to other non-occupational environmental sources of dusts such as soot from coal combustion from cooking and/or heating or lived near nuisance dusts or soil erosion dust. This hypothesis is supported by studies of quarrymen5 and some air pollution 6 7 studies. Pham et al propose an interesting mechanism for the carcinogenesis of general dust exposure in their efforts to explain the high rate of stomach cancer in iron ore miners. They suggest that the inhaled iron oxide dust may have acted by production of free radicals arising from surface activity by a mechanism of oxygen reduction which is greatly increased in freshly generated dust. This physicochemical surface activity has also been found in silica and asbestos fibre dust (Costa et al 1989).

It may also be relevant in the high rates of stomach cancer found in those in grinding occupations in bearing plants.

3. Gastric cancer risk may be associated with exposure to nitrosamines, their precursors, or to nitrosation facilitators. Nitrosamines are formed by the union of nitrites with other amines, therefore any factor that increases exposure to nitrites, nitrates or their precursors would increase the risk of gastric cancer8. Coal miners were exposed to nitrosamines and nitrogen oxides released from coal by pyrolysis; nitrates from inhalation of particulate after blasting (nitroglycerin or ammonium nitrate); dietary sources in smoked or salted fish or bacon (smoked foods also contain benzopyrene); drinking water sources from wells contaminated with high levels of nitrates (1981 studies by Armijo et al suggest a co-factor of arsenic and selenium to explain nitrate exposure and gastric cancer); nitrosation enhancement from chewing tobacco and/or bacterial contamination from sewage polluted water, poor dental health, and poor preservation of and storage of food. Similarly nitrosamines may be implicated in stomach cancer in rubber workers (Nutt et al) - exposure was very high when N-nitrosodiphenylamine was in use. Nitrosamines are also found in soluble oil and synthetic cutting fluids and may be relevant to the high stomach cancer rates found in bearing plant workers.

Another hypothesis was proposed by Meyer et al9 to explain stomach cancer in coal miners but was later refuted by Ames and Gamble10 and not supported by subsequent studies11. Meyer et al suggested that stomach cancer risk is increased when particle retention in the lungs is decreased, ie particles are more rapidly cleared to the mouth and swallowed when there is no pulmonary impairment; gastric cancer risk would be lower when lung clearance is reduced because of pulmonary function impairment.

These hypotheses are of interest and could potentially explain a high proportion of the observations of increased mortality from stomach cancer in various occupational groups. Alternative mechanisms are obviously possible and would be relevant to explain the Kusiak et al suggestion that exposure to chromium is the main cause of the high rate of stomach cancer among Ontario gold miners12 observed in that study.

4 REVIEW OF THE LITERATURE

4.1 METHODS

A Medline search covering the period 1966 to 1993 was carried out in April 1993 using the following search strategy
[occupational diseases (MH) or occupational exposure (MH) or occupational health (MH)] and [stomach neoplasms (MH) (includes gastric cancer MH)].

Fifteen articles had been identified and sent by the IDSP with the initial request for the review. A further 73 relevant articles ie review articles or those describing an epidemiologic study linking occupation and stomach cancer were identified in the Medline search. Some were retrieved by the IDSP. A research assistant was hired to retrieve the remaining relevant articles and a total of 82 articles were retrieved and reviewed for further study. One review article was retrieved from the personal files of the reviewer. Those which met the following criteria were selected for more detailed analysis:

  • case control stomach cancer incidence or mortality studies which sought a link to occupations with a potential exposure to dusts or metals
  • cohort studies of occupational groups exposed to dusts or metals in which stomach cancer was a health outcome studied
  • reports of studies in which the methods were described adequately to allow for critical appraisal of the study
  • review articles addressing the relationship between stomach cancer and occupation.
  • Studies in which the main exposure identified was asbestos or ethylene oxide were excluded.

    Three previous reviews were identified, one13 a general review which described existing studies without critically evaluating them, a review of stomach cancer in the rubber industry14, and a review of lung and stomach cancer in gold miners published by the IDSP15. Forty-two of the primary etiologic studies which met the above criteria were reviewed in detail.

    These studies were evaluated with a view to identifying significant threats to the validity of the conclusions based on methodologic flaws relating to the study population, the controls, measurement of exposure, ascertainment of the stomach cancer outcome, accounting of confounders, or the adequacy of the study design. Studies were reviewed for methodologic quality but not systematically rated or ranked. A summary table describing the studies organized by occupational group was developed. Where there was more than one published paper or report of a study or its update all the relevant reports were identified in the references but only the results and methods of the most recent update was described in the summary table. No attempt was made to quantitatively rate the studies or rank them. The methodologically strongest studies of those reviewed are referred to in the discussion of Section 5.

    4.2 EPIDEMIOLOGIC CONSIDERATIONS

    The following methodologic issues were taken into account in the reviews:

    A Study Design:

    The studies dealing with occupation and stomach cancer are generally either cohort studies or case control studies. In general cohort studies are a stronger design with less risk of serious bias than case-control studies.

    Cohort studies include:

    1) prospective cohort studies:

    A) prospective cohort cancer incidence studies, ie a study which identifies a cohort of exposed people at the time the study is being conducted and then follows those workers prospectively identifying any all cases of stomach cancer and comparing this observed incidence rate to that expected for a similar population of the same age and sex not exposed to the exposure under study;

    B) prospective cohort cancer mortality studies, ie a study which identifies a cohort of exposed people at a point in time and then follows those workers prospectively identifying all deaths from stomach cancer and comparing this observed mortality rate to that expected for a similar population of the same age and sex not exposed to the exposure under study;

    Prospective cohort studies are the strongest design methodologically because the information collected for them has a greater likelihood of being relevant and accurate rather than depending on existing data bases designed for other purposes. An entire cohort can be identified from an existing and living working population. Exposure information can be measured in the present and subjects can be directly questioned about their personal exposures in the past and present and about potential confounders prior to the development of health outcomes and these factors can be followed over time. Such studies are extremely rare because they are very costly and for cancer studies must be continued for 25 to 40 years to allow for cancer latency and deaths to occur.

    The only study that was identified by the researchers as a prospective mortality study was that of the Viennese workers (Neuberger et al 1986; 1990). In this study 247,064 workers in 1,089 Viennese workplaces were examined and questioned (and x-rayed if dust exposed) by a mobile occupational health team in the period between 1950 and 1960. Decades later the mortality experience of 1630 male workers from this cohort found on the initial occupational history to be exposed to non-fibrous dusts was followed. Because the initial survey was not undertaken and designed specifically as a prospective mortality study on stomach cancer it in fact has some of the weaknesses of a retrospective cohort study and would probably be better described as a cross-sectional study repeated twice on the same population.

    2) Retrospective cohort studies

    A) Retrospective stomach cancer incidence studies

    In these studies a population of exposed workers is identified from past records and then the members of this population are cross-linked with a regional or national cancer registry which records all new cases of cancer. The rate of cancer in this population standardized for age, sex and calendar year is then compared to that of the regional or general population or sub groups within the population who vary in degree or nature of exposure are compared. The strength of such studies is dependent in part on the accuracy and completeness of the cancer registry. In general Scandinavian registries are believed to be much more complete than American ones.

    Cancer incidence studies have a slight advantage over mortality studies in that all cases of the cancer under study are included whereas in a mortality study only those cases which die of their cancer are included. In a study in which a high proportion of the study cohort is still alive at the end of the follow up period there may be considerable differences in the results of an incidence study versus a mortality study. Stomach cancer generally has a low survival rate and therefore this is less of a concern than with some other cancers such as colon or laryngeal.

    B) Retrospective stomach cancer mortality studies

    In these studies a population of exposed workers is identified from past records (employer payroll, union membership, pension, national health service or other similar records) and then the members of this population are cross-linked with a regional or national vital statistics information. Frequently death certificates are sought for all members who are identified as deceased. The rate of stomach cancer in this exposed population standardized for age, sex and calendar year is then compared to that of the regional or national general population and/or sub groups within the population who vary in degree or nature of exposure are compared. While more efficient and less costly than a prospective study, these studies may have difficulty reconstructing the occupational/job history of the study group and therefore have inadequate exposure information; employment records may only go back 30 or 40 years but exposures were highest 50 to 90 years ago, eg mining populations. This method is dependent on complete and accurate government vital statistics and diagnosis on death certificates. Accuracy of such information may be better in recent decades than in previous ones, or may be better in some regions of a country than others.

    Sometimes in such a cohort study all the cases of stomach cancer will be used to carry out a nested case-control study. The exposures of the sub-group of stomach cancer cases will be compared to a control group taken from the same occupational cohort but who did not develop stomach cancer in order to identify exposures associated with stomach cancer. Nested case control studies allow more detailed analysis of exposure and/or other risk factors than were evaluated for the entire cohort. The main difficulty in such a study is identifying an appropriate control group particularly if the exposures of interest are also associated with other health outcomes.

    3) Proportional mortality and proportional incidence studies

    In these studies one cannot identify the entire cohort at risk but can identify those who have died or those who have developed cancer. In the stomach cancer studies this usually involved reviewing a regional death certificate data bank in which occupation, cause of death and other demographic information are available or death records available from company or union pension records. One can therefore identify a cohort in a particular occupation or industry and look at what proportion of this cohort have died of stomach cancer (the Proportional Mortality Rate, PMR) compared to what one would have expected in the general population of that region standardizing for age, sex, race and year of death. In stomach cancer, proportional incidence studies usually define an occupational cohort from a cancer registry and calculate an SPIR (standardized proportional incidence rate) of the proportion of those who developed stomach cancer as a proportion of all cancers compared to the remaining members of that registry. Proportional mortality and incidence studies are inferior in design to other cohort studies because the deaths included may not be representative of all deaths if the full cohort had been included and followed. Also when the PMRs for several diseases are elevated counterbalancing proportionate mortality deficits will occur for other causes because we are dealing with proportions and not absolute rates and by definition the total number of observed deaths from all causes will equal the expected number. If an exposure is associated with several types of cancers or several other causes of death (eg stomach cancer, lung cancer and non-malignant respiratory disease) the PMR of a single cause, eg stomach cancer, may underestimate the SMR. In PMR studies one needs to ensure that the healthy worker effect is similar in the study outcome as in the control outcomes. Some consider PMR studies as a variant of case control studies and others argue that they are of lower methodologic quality because they are at greater risk of selection bias than case control studies.

    Case-control studies

    Here cases of stomach cancer in a region are identified, usually from cancer registries or from several hospital pathology departments who agree to participate in a study. The exposures of these cases are then compared to a control group of other patients, either other cancer patients, patients with other diagnoses unlikely to be associated with the exposures studied, or a community group matched for age, gender, date of diagnosis, and/or neighbourhood. Frequently the exposure information is very poor and is not sought directly from the cases and controls but rather is inferred from occupational information in the registry or hospital record and is therefore of limited accuracy or only reflects occupation at the time of diagnosis not lifetime exposures. (A very big exception to this is the Siemiatycki study in which great care was taken to elicit full occupational histories from incident cases.) There is much greater risk of selection bias and information bias (especially if the information on exposure is taken after the diagnosis is made) in case control studies. There is tremendous difficulty ensuring that the control population is appropriate, ie does not differ from the cases in ways that may be relevant to the exposure under study. Recall bias in those with disease compared to community controls is a major limitation of these studies. Therefore these studies are generally inferior to cohort studies.

    B Population

    There are a number of concerns about study populations that can bias results. With cohort studies the main concern is that all potential members of the cohort are identified and their vital status and/or cause of death is accurately ascertained during the follow up period. If a high percentage of the cohort are either not identified at the start because of incomplete records or information about their vital status, cause of death or job history is not available the subjects studied may not be representative of the cohort as a whole. Also the methods of ascertaining vital status and cause of death may vary enormously in their accuracy and must be evaluated.

    In some studies identified as retrospective cohort mortality studies (eg Finkelstein et al, 1987; Meijers et al 1991) the occupational cohort followed is identified on the basis of another disease eg a pneumoconiosis registry. Thus the full cohort of exposed workers is not identified but a sub-set of workers. It is important to determine whether the subset is representative of the full exposed cohort. For example, are those who develop pneumoconiosis somehow different in terms of risk or do they represent the sub-group with high exposure to silica, coal or other mine dusts? Interestingly, in both these studies of presumably very highly exposed workers the association between stomach cancer and occupation is very high - an excess of three to four-fold.

    Another concern in cohort mortality studies is the capacity of the study to detect increases in cancer rates in the study population. This is dependent upon the total number of workers studied, the number of years since first exposure among those who have died (adequate latency), the number of years members of the cohort were exposed, their age distribution (are there sufficient older workers, ie those over 60 or preferably over 70, to detect an excess of cancer deaths?), and the proportion of the cohort who have died at the end of the follow up period. This last factor is often ignored. If only 25% or less of the cohort has died one may not pick up excess cancer death rates which would be evident when a majority of the cohort has died. This factor may be important in stomach cancers in particular because, in the general population, these cancers tend to occur in later life. The studies of asbestos-exposed US and Canadian pipefitters by Selikoff et al demonstrated that excess rates of mesotheliomas and cancers often occurred 35-40 years after initial exposure. Few of the cohort studies reviewed described the latency or age distribution of their cohort population. This was an important concern in the Ontario miners' study (Kusiak et al; Muller et al) in which the total miners' cohort included many different groups of miners who varied considerably in the age distribution of the members of the subgroups, the number of years exposed prior to and during follow up and the proportion of the sub-group who had died by the end of follow up. Although a very high number of workers were included in this study population and the number of person-years observed was very high, it is not clear what proportion of workers had latencies of greater than 20 or 30 years. If a high proportion of those who died had latencies of less than 30 years or if a high proportion of those in the study population with the longest latencies have not yet died at the time of the mortality follow up, the capacity of the study to detect excess stomach cancer deaths will be reduced. This problem is usually avoided in PMR studies because all those studied have died.

    Another concern in cohort studies is how to deal with the healthy worker effect. In many of the cohort studies reviewed here the overall mortality of the cohort and the cardiovascular disease death rate had a statistically significant deficit, usually an SMR of about 80 when compared to the national population mortality rates. Some researchers believe other cause specific death rates should be adjusted for the healthy worker effect. Thus an unadjusted SMR of 1 may suggest an excess of, for example, 25%. The healthy worker effect in these cohorts may thus lead to a systematic underestimate of stomach cancer risk. Others interpret cancer SMRs of 1 as reflecting a lack of a healthy worker effect with respect to cancer. The issue is far from clear.

    C Controls

    The main concern with control groups is that they be as similar as possible to the cohort with respect to potential confounders such as age, sex, diet, socioeconomic status or immigrant status or that differences be measured and accounted for in the analysis. This is especially a concern in case control studies where unaccounted for factors may influence the likelihood of exposure in the control group. In cohort studies the concern most frequently raised was the appropriateness of national versus local mortality or cancer incidence rates as comparisons. There appears to be considerable regional variation in stomach cancer rates and frequently the local stomach cancer rate was higher than the national rates. Is this a function of a large exposed group locally or differences in diet or other confounders? If it reflects occupational exposure then the national comparisons would be most appropriate; if it reflects the presence of local confounders then local rates may be more appropriate. Again resolving this issue is not easy.

    D Exposure

    The lack of exposure measurement was the greatest limitation of all the studies reviewed. None of the studies had direct measurements of a specific exposure or occupational hazard. All relied on occupation as the primary measure of exposure. Some studies provided weighted exposure measurements based on years in a particular occupation or job taking into account latency factors. Most relied on existing data bases for occupational history. Only the study by Siemiatycki et al took life time occupational histories directly from workers. The Viennese Dusty Worker Study (Neuberger et al) took an occupational history from the worker at the initial assessment and assumed exposure remained constant in the intervening years through to follow up. The PMR studies of US bearing plants (Park 1988, et al; Silverstein et al, 1988) did examine the working conditions at the plants, reviewed previous hygiene reports and had records of the duration that individual workers spent in various departments or job categories and thus were able to hypothesize various potential exposures. Similarly in the Ontario gold mining studies investigators had access to hygiene information and the types of exposures that existed in various jobs over time. But this type of information was not taken into account in any quantitative way to test hypotheses about specific exposures. Thus while a number of studies demonstrated a strong relationship between stomach cancer and occupation, none attempted to show a direct association between stomach cancer and a specific occupational agent.

    E Health Outcome

    The health outcome of interest in these studies was either the incidence of stomach cancer or stomach cancer mortality. In general the stomach cancer incidence data relied on cancer registries which usually base their diagnoses on pathology reports and thus are more likely to be accurate than death certificate information. Stomach cancer may not have always been appropriately diagnosed particularly in previous decades. In the past it was not uncommon in some regions to confuse stomach cancer with mesothelioma, oesophageal or other abdominal cancers especially if diagnosed after metastatic spread.

    In some studies (eg Hamburg gas workers, Berger 1992) different methods of ascertaining cause of death were used in the same study population, eg hospital records, family doctor files, and/or statements of family members which may vary in accuracy, and are definitely different than the methods used to ascertain death in the control population.

    Another bias can be introduced when cases are identified in occupational cohorts being followed for respiratory disease and malignancy who are thus more likely to have post mortem examinations and a subsequent diagnosis of gastric cancer. This may well have been the case in the Welsh coal miners study (Atuhaire et al, 1986) where non-miners were more likely to have a diagnosis of carcinomatosis, site unspecified with no post mortem compared to miners with pneumoconiosis pensions who frequently had post mortems.

    Confounders

    As mentioned in the introduction a number of non-occupational factors have been identified as associated with stomach cancer. These include diet (eg foods that are smoked, salted, contain nitrates, nitrites), low socioeconomic status, smoking (studies are contradictory on this point and smoking is, at best, a weak predictor of stomach cancer), ethnic origin (eg East or Central European), immigrant status, northern vs southern areas, chewing tobacco; older age; exposure to non-occupational radiation; exposure to other known carcinogens other than the occupational exposure being studied. Studies varied with the degree to which these factors were addressed. But several studies did control for age, diet, socioeconomic status, and ethnicity or immigrant status and still showed strong relationships between stomach cancer and occupation.

    The role of ethnicity in stomach cancer is confusing. A wide number of ethnic groups have been implicated. It may be that such studies are confusing immigrant status with ethnicity, ie the risk of stomach cancer may be higher in immigrants who may have lower socioeconomic status, poorer diets and/or use coal for cooking and heating rather than due to ethnicity per se.

    4.3 STUDIES REVIEWED

    The studies reviewed are presented in Table 1.

    4.3 TABLE 1  SUMMARY OF EPIDEMIOLOGIC STUDIES OF GASTRIC CANCER AND OCCUPATION
     
    AUTHOR                          STUDY DESIGN                 STUDY POPN                        CONTROL POPN                      CONFOUNDERS                                 RESULTS:                             
    OR/SMR/p-value
          
        COMMENTS                  
    MINERS/GAS PLANT WORKERS
    Kusiak16 17 18 19 20     retrospective cohort mortality     54,128 Ontario miners (gold, nickel, copper, uranium, and other ores) followed by provincial chest clinics after 1955 who died between 1955-87 (exposed to arsenic, chromium, mineral fibre, diesel emissions, aluminum powder)     Ontario mortality data     age, gold vs non-gold miners, born in N Am or not     Gold miners SMR 152 (CI 125-185) for stomach cancer; no increased SMR for non-gold miners (uranium, copper, nickel) except those born outside N Am (SMR 160 95% CI 105-233);
    -gold miners born in N Am SMR 133; born out of N Am 177; SMR greater in those who were <60yr (167 vs 145 miners 60-74) and born out of N Am (160); association statistically significant between
    gastric cancer mortality and time weighted exposure to dust in gold mines, chromium (but not arsenic or mineral fibre)
        no mention of denominator figures ie total number of deaths for each group of miners; no direct exposure measurements; difficult to assess power to detect increased stomach cancer in various gps of miners without knowing total number of deaths in each gp
    Hodgeson21     retrospective cohort mortality study     3072 UK (2 Cornwall mines) tin miners who worked at least 12 months between Jan 1/41 and May 1/84 excluding those born before 1880 or over age 60 when began mining exposed to radon, silica, arsenic (13 miners);(851 deaths)     national (England, Wales) mortality rates     calendar period, age, total underground exposure, time since last exposure     SMR Cancer of stomach 1.41 n.s. (27 cases)     only 28% of cohort had died during the study period
    Lawler22 retrospective cohort mortality study     10,403 white male Minnesota iron-ore miners (45% underground, 55% above ground) employed for at least 12 months prior to Jan 1/66 followed from Jan 1/35 to Dec 31/78 (4699 deceased, 4478 death certificates obtained)     US white males mortality rates; St Louis county mortality rates     age, calendar time, above vs underground mining     total cohort SMR stomach cancer 172 (p<.01); underground miners SMR 167; above ground 181, if hired 1937-65 SMR=223, if native born SMR=155; but SMR=101 if controls=St Louis county; foreign born underground SMR=186, foreign born above ground=234; when compared to St Louis county no statistically signif increase except for Finnish born miners SMR 173
     
        differences between underground and above ground miners (underground started exposures earlier, worked longer, >3x more were foreign born 39.3% vs 11.7%)
    Atuhaire23     30 year prospective (?) mortality cohort study     7939 male miners and non-miners in the Rhondda Fach, Wales followed for 30 yrs     death rates for England and Wales for the same period; comparison between miners, ex-miners and non-miners     age, time period     Miners/non-miners SMR=148 p<.05); non-miners SMR=131 (p>.05); no statistically signif differences between miners and non-miners
     
        % who died of total cohort not stated; very few non-miners were age>45 (404/2000=20%); only 926 of 6182 miners were over age 65
    Swaen24     case control gastric cancer incidence study     323 male cases of gastric cancer diagnosed between Jan 1/73-Dec 31/83 in southern Netherlands
     
        323 controls from same pathology dept matched for date of birth within 30 days     age, coal mining or not     no signif difference between cases and controls with respect to occupation as a coal miner     very low numbers, ?appropriateness of controls
    Swaen25     case control gastric cancer incidence study     683 male cases of gastric cancer diagnosed between Jan 1/73-Dec 31/83 in 5 pathology depts in southeastern Netherlands
     
        683 controls diagnosed in the same pathology dept matched for date of birth
        age     OR 1.15 (95% CI .89-1.47)     as above
    Meijers26     retrospective pilot cohort mortality study     165 deaths of 334 Dutch coal miners randomly selected from 5400 Dutch coal miners diagnosed as having pneumoconiosis in routine cross-sectional x-ray examination between 1956-60 and registered with the General Mining Fund followed to Dec 31/83
     
        cause, age, calendar time specific mortality rates for total Dutch male popn     age, time period, smoking     SMR stomach cancer =401 (95% CI 2.3-6.5)     not based on a full occupational cohort; high mortality rate (49.4%), long follow up period (52% had >30 yr employment, 87.5%>20 yrs)
    Pham27     prospective proportional mortality study     13,801 French iron miners from the Lorraine basin alive on Jan 1/82 followed from Jan 1/82-Dec 31/86 (1,283 deaths ie <10% cohort); <20% cohort >60 yr     proportion of deaths of French male popn     age, calendar time, smoking     PMR stomach cancer 2.31 (p<.001); surface workers PMR 2.18 (p<.05); underground workers PMR 2.37 (p<.001); underground workers working 1-19 yr PMR=2.63, 20-29 yr PMR=2.66 (p<.001), >30 yr PMR= 2.01 (p <.01); (51 deaths of stomach cancer; 80% of those who died of stomach cancer were 60-80 yr old)     limitations of PMR study; small % died; relatively young pop'n for mortality study esp for stomach cancer; multiple methods of ascertainment of cause of death of study population
    Weinberg28     case control     228 deaths 1978-80 for 4 counties in Penn (from state vital stats with death certificate after gastric cancer identified in v.s.
    178 next of kin interviewed by tele (107 males, 71 females)
        3 controls/case (107M/71F digestive cancer, 107/71 ASHD death, and 80/58 live neighbourhood control (dig-same sex, race, county, 5-yr age; ASHD- next death following case of same sex, race, county, within 5 yr age)     tobacco chewing, smoking, age, SES (neighbour-hood; income), ethnicity (East European, foreign born); gender     East European origin OR:
    1.57 dig cancer
    2.05 ASHD
    1.88 neighbourhood;
    occupation n.s. increase in coal miners and farmers and wives of c miners; gas cooking protective
        poor study; info from next of kin, SES info poor; death certificates requested for cases only; neighbourhood controls problematic; data on occupation poor; power low
    Berger29     retrospective cohort mortality study     all 4908 male Hamburg gas plant workers employed for 10+ yrs between Jan 1/00 - Dec 31/89 who died between 1952-1989 (2699 total deaths); sub cohorts: 789 gas furnace workers with heavy exposure to coal tar gas (572 deaths); 3401 other blue collar workmen in other parts of the plant (1708 deaths)     calendar period, age, and cause specific death rates for German males; subcohort of 718 white collar workers internal controls (419 deaths)     age, calendar period, smoking     SMR stomach cancer gas workers = 177 (95% CI (120-251); other workmen SMR=163 (88-142); white collar workers SMR=57 (27-105); when gas workers compared to white collar workers as standard SMR=284 (95% CI 205-385); smokers had higher SMR compared to non-smokers; strong healthy worker effect for most other causes of death
     
        multiple methods of ascertainment of cause of death of study pop'n including asking family members; no controls of ethnic identity, immigrant status, SES

    WOOD/PAPER DUST
    Siemiatycki30     case control     156 Montreal incident cases of stomach cancer from multiple hospitals     1524 cases of other cancers     age, ethnicity (French or not), SES, smoking, salubrity of work history     OR 1.7 for exposure to wood dust with 5 stratifying variables (n=35) (95%CI 1.1-2.8)     well designed study (poor wrt ethnic variable); limits of exposure coding
    Milham31     Proportional mortality study     2689 deaths of active or retired union members identified from listings in the Pulp, Sulfite and Paper Mill Workers' Journal 1935-64 (2113 death certificates found, 79%); 68 stomach cancer deaths     age, calendar year specific US proportional mortality rates     age, calendar year     PMR stomach cancer = 176 (p < .05)     significant methodologic problems; no attempt to determine what proportion of total cohort deaths included in journal listings; Canadian workers made up 25-30% but treated as US in PMR calculations
     
    Roscoe32     retrospective cohort mortality study     2294 white male wood model makers who had worked at least one month at one plant of each of the 3 major US auto makers between 1940 and 1980, followed 1940-84; 706 deaths (31%)     Wayne County, Mich mortality rates; US mortality rates     age, calendar yr, place of birth     overall stomach cancer compared to national mortality rates 1940-80 SMR = 1.6 (95% CI 0.9-2.6), (17 cases); using Wayne County mortality rate SMR 1960-80 1.6 (0.9-2.7), (15 cases); US mortality rate comparison 1960-80 SMR = 2.2 (1.2-3.6);     difficulty with exposure measures in case control study
    Svirchev33     Proportional mortality study of BC Wood and Pulp workers     BC wood and pulp workers aged 20-65, 1950-78     BC mortality data     not stated     PMR stomach cancer carpenters 155 (95% CI 126-191); cabinet makers 150 (40-384); woodworkers misc 179 (89-320; pulp mill workers 115 (42-250); no increases for other wood/pulp workers
     
        methods not described in this paper (letter)
    Olsen34     Proportional incidence ratio (SPIR) study     cases of stomach cancer notified to the Danish Cancer Registry 1970-84 for ages 16-66 employed after Apr 1/64 linked to personal employment histories on file in a nationwide Pension Fund for those in wood working occupations
     
        proportion of stomach cancers among all Danish employees     age (5 yr gps), time (1 yr periods)     SPIR stomach cancer 204 (p<.01) basic wood industries; 102 wood manufacturing industries; 139 (p<.01) agriculture, forestry, fishing; 101 manufacturing     study unable to detect cancers in former Danish workers >age 66; limits of PMR
    Toren35     case control     33 men who died between 1961 and 1985 of stomach cancer from 3 parishes near 2 paper mills     550 men over age 40 who died in the same parishes of non-malignant non-respir diseases (no indiv matching
     
        age     no statistically signif increase risk of exposure among stomach cancer cases     very low power, low numbers
    SILICA/NON-SPECIFIC DUSTS/FOUNDRIES/CEMENT PLANTS
    Finkelstein36     retrospective "cohort" mortality study (see comments)     276 Ontario silica exposed non-miners in various industries (with 212 deaths - 204 death certificates located) and unspecified number of Ontario miners (with 905 deaths) identified from a WCB cohort of workers awarded compensation for silicosis between 1940 and 1985
     
        age, sex, time specific Ontario mortality rates     smoking, decade of initial award, industry type, birthplace     miners with silicosis SMR stomach cancer 188 (CI 120-290); silica exposed non-miners SMR stomach cancer 366 (CI 147-754)     high proportion died; limitations of accuracy of WCB files; not a true historical cohort study in which cohort is identified on the basis of exposure but rather mortality study of a cohort of cases for another diagnosis which suggests high silica exposure
    Coggon37     case control stomach cancer incidence study     95 cases of stomach cancer newly diagnosed in pathology reports or in the endoscopy, radiology and surgical depts of the local hospitals of Stoke-on-Trent and Newcastle-under-Lyme county districts in those less than 70 yrs of age between Jan/85 and Aug/87
     
        2 controls/case (190) matched for age (within 2 yrs), sex from the county district popn registered with NHS GPs     age, sex, diet (salad vegetables, fruit, salt)     Relative Risk (RR) of cases holding a manual job in one of the 4 local dusty industries=1.4 (95% CI 0.8-2.4); RR manual work in rubber industry controlling for diet=2.5 (95% CI 1.0-6.4) and when controlling for social class as well RR=3.3 (1.2-9.2)
     
        low numbers; minimal exposure information; no accounting for duration of exposure
    Neuberger38 39     prospective cohort mortality study     1630 male workers in multiple industries > 40 yr exposed to silica and "inert" dusts screened by the Vienna Occupational Health Care Unit in the 1950's followed to Dec/1985 by death certificates (87% had died)     death rates in general Viennese population; internal control: 1650 male Viennese workers without occupational dust exposure (84% died) selected from the same screening examination records matched for age (yr of birth), yr of first examination, smoking status at time of first exam;
     
        age, smoking     SMR stomach cancer in all exposed=166 (p<.01); in foundry workers 177 (p<.01); other metal workers 158 (p,.05); Internal controls SMR=90     occupational history limited to occupation at time of first examination; very crude exposure information; no way of ensuring that internal controls not exposed during subsequent employment (bias here would have underestimated association)
    Bross40     case control cancer incidence study     14,000 cases of cancer admitted to Roswell Park Memorial Institute, NY 1956-65 with occupational histories taken     patients admitted in the same period with a non-neoplastic disease and patients employed as clerks     sex, age, yrs in occupation, smoking, country of origin     For stomach cancer RR of Carpenters >2 (p <.05); shoemakers RR=16.5 (p<.05); millwrights RR=5.5 (p<.05); when only those born in US included the following occupations RR>2, p<.05: blacksmiths, brickmasons, lumber industry, millwrights, paper industry operators, textile industry operators
     
        limits of case control study; crude occupational information
    Sorahan41     retrospective cohort mortality study     10,491 male UK steel foundry workers who worked for 1 or more yrs 1946-65 and followed 1946-85 (3110 died, cause known, 29%)     mortality rates for England and Wales     age, sex, calendar year, duration of employment, age at starting employment, factory, yr of follow up     All steel foundry workers SMR stomach cancer = 137 (p<.01); furnace workers SMR=265 (p<.01); furnace repair SMR=359 (p<.001); highest SMR's for those working 10-19 yrs
     
        only 29% of cohort dead by end of follow up, therefore difficult to look at dose response relationships
    McDowell42     retrospective cohort mortality study (SMR)     607 men known to be cement workers in 1939 in north Kent, UK, and alive in 1948 identified by reviewing National Health Service Central Register records followed from Jan/1948 to Dec/1981 (419 deaths, 22 stomach cancer deaths)
     
        male mortality rates for England and Wales     age, social class, calendar yr of death     Stomach cancer SMR=175 (p<.05); packing or loading labourers (high dust exposure) SMR=321 (p>.05); SMRs above that for stomach cancer in social class IV or V     69% of cohort had died; no information on duration of employment or level of exposure except by job title; ? accuracy of occupational information
    Amandus43     retrospective cohort mortality study (SMR)     4231 white male workers employed in 23 US non-asbestos cement plants for at least 5 years between 1950-1960 (5292 men met inclusion criteria but lack f vital status or incomplete occupational or other history on 1061; death certificates available for 31% of cohort (32.6% cohort died; 135 stomach cancer deaths)
     
        age-year specific US white male mortality rates     age, calendar year, latency intervals, years of employment, country/continent of birth (US/non-US)     SMR for stomach cancer overall 135 (not statistically significant); for those with >20 yr latency SMR=141 (p>.05); for those born <1900 followed up 1965-9 SMR=491 (p<.05), those born >1900 and followed 1965-9 SMR=221 (p>.05); for other periods of follow up (<1964, >1970) no signif SMRs     low power - negative results unconvincing because of study limitations; less than a third of cohort had died; high proportion of original 8714 workers enumerated excluded: 1791 because began working before 1950, 828 excluded because work history before 1960 unavailable - exclusion of more older workers limiting potential of finding a statistically signif increased SMR; no attempt to check accuracy of stomach cancer diagnoses particularly in early decades of study
     
    Wu-Williams44     case control     137 incident male cases of stomach cancer under age 55 identified through the LA County Surveillance Program (cancer registry) interviewed between 1975-1984     137 age-race-sex matched neighbourhood controls     social class, nativity, dietary factors, smoking, alcohol     Metal dust exposure RR (conditional logistic regression) =1.7 (95% CI 1.0-3.0); metal dust exposure 1-19 yr RR = 1.4 (0.7-2.5); 20+ yr exposure RR= 7.4 (1.6-33.5)     Despite low numbers and controlling for other confounders occupational metal dust exposure show very high association with stomach cancer; large number of cases' occup info came from surrogate
     
    JEWELLERY WORKERS
    Dubrow45     PMR study of Rhode Island jewellery workers     1334 female and 1807 male RI jewellery workers listed in RI death certificate records 1968-78     all other decedents listed in RI death certificate records 1968-78     standardized PMR for gender,race (B,W), age, yr of death     PMR 174 (CI 114-265) among female jewel. workers; neg for males (different jobs)     limitations of PMR method, of death certificate data; exposure based on only broad occup categories not specific exposures; no control for SES, immigrant status or ethnicity
    Sparks46     PMR study of Attleboro, Massachusetts jewelry workers; nested case control     931 male who died in Attleboro, Mass. between 1956 and 1975 and were listed as jewelry workers on death certificates; 14 stomach cancer deaths     US white male mortality data; case control internal comparison: all non-stomach cancer deaths     age, calendar year, ethnicity (French Canadian)     Overall PMR for stomach cancer = 1.2; PMR for polishers = 3.9; Odds ratio comparing stomach cancer to all other causes of death in polishers = 4.4 (p<.01)
     
        Small numbers; only one p-value cited - ? statistical significance of other results; weak study
    SHOE/LEATHER INDUSTRY
    Garabrant47     PMR with nested case control     PMR: 78 female and 758 male leather workers; 767 F and 1195 M shoe workers in 3 Massachusetts communities listed in Mass death certificates 1954-74     all non cancer decedents listed in the above death certificates of 3 Mass. communities     standardized PMR for gender, age, yr of death     Male leather workers PMR stomach cancer 1.69 (CI 1.04-2.73); F leather PMR 2.80 ns (2 obs deaths); case control study did not show statistically signif increase in exposure in those with stomach cancer
     
        limitations of PMR method, of death certificate data; PMR exposure based on only broad occup categories not specific exposures; no control for SES, immigrant status or ethnicity
    Edling48     case control cancer mortality study looking at exposure to tannery occupation     19 male cases of stomach cancer were identified between 1951-1979 in a local parish death registry in SE Sweden which had a very large tannery between 1873-1960
     
        388 male controls who had died of causes other than cancer during the same period in that parish     age     SMR for stomach cancer and exposure to tanning (as occupation on death registry) = 1.5; Mantel-Haenzel rate ratio point estimate = 1.6 (95% CI 0.6-4.0)
     
        low numbers, low power; plant closed 1960, those who switched jobs after that may be listed in registry with another occupation
    BALL BEARING
    PLANT
    Park49     PMR mortality study of a employees from a Connecticut ball bearings plant; case control study for specific disease outcomes and exposure categories (exposures to grinding dust, abrasives, water-based cutting fluids and straight cutting oils)     PMR: all hourly employees with 10 or more yrs of service who died between Jan 1/69 and July 31/82 = 768 deaths (702 had death certificates & job histories - 565 white males, 20 non-white males, 116 white females, 1 non-white female)
    Case control: stomach cancer deaths from above pop'n (11 stomach cancers in white males)
        PMR: US proportional mortality rates
    Case control: all deaths excluding other GI cancer sites, lung cancer and non-malignant GI disease
        age, sex, race, calendar yr, Central European origin     PMR: for stomach cancer in white males PMR=3.1, p<.002; females PMR=3.1, p=.15; men in grinding PMR=3.8, p=.006; grinders with 30+ yr latency PMR=4.2, p=.03);
    Case control: crude mortality OR for stomach cancer and grinding=5.9, p=.06; water-based cutting fluids OR=6.2, p=.05; no association found for straight oils
        limits of PMR analysis; attention to exposure more detailed than most studies; good discussion of threats to validity of this study; case control results suggest stomach cancer risk in grinding is likely due to additives in fluids not present in straight oils (eg petroleum sulfonates, amine soaps, nitrite corrosion inhibitors) abrasive aerosols, or oils with abrasive debris
    Silverstein50     Proportional mortality study, nested case control     PMR: 1532 white male bearing plant unionized workers whose year of termination was 1950 or later and who died between Jan 1950 and June 1982 (exposures to cutting fluids, abrasives, grinding dust);
    Case control: 35 stomach cancer cases
        PMR: national US proportional mortality rates;
    Case control: all non-cases of stomach cancer
        age, yr of death, Central European origin, French Canadian origin     PMR: overall PMR for stomach cancer = 1.97 (p &lEt; .001); PMR for those with 10 or more years exposure to grinding (comprehensive) = 3.39 (p < .001); PMR Tool grinders = 5.2 (p=.04); no excess found in machining occupational categories;
    Case control: Crude Mortality Odds Ratio 10 or more yrs Comprehensive grinding 2.37 (p=.03); logistic regression controlling for non-US origin Adjusted Odds Ratio for comprehensive grinding = 2.25 (p=.024)
     
        Study appears to have begun as a retrospective cohort mortality study but been analyzed as a PMR study. Cause of death and job information available on 87% of 2027 deaths identified - 1532 in white males; exposure data based on years in specific job groupings with knowledge of hygiene values - more sophisticated exposure estimates than most other studies
    RUBBER WORKERS
    Sorahan51 52     retrospective mortality cohort study     36,691 male UK rubber workers in 13 factories who worked for minimum of 1 yr 1946-60 followed 1946-85 by the National Health Service Central Register or Scottish General Register Office (11,765 dead, 32.1%; 359 stomach cancer deaths); internal controls of different occup gps     mortality rates of England and Wales or Scotland     age, calendar yr, duration of employment, factory, yr of starting employment, age at starting employment, yr of follow up     SMR stomach cancer= 113 (p<.05); age<50 yr SMR=139; age >50 yr SMR=110; greatest increases in SMR for those with latency of 10-19 yrs; SMR higher for those first employed 1956-60 (SMR=133, p<.05) than for those first employed 1946-50 (SMR=117, p<.01); suggestion that risk associated more with exposure to dust than fumes or solvents but test statistic not statistically signif
     
        exposure data limited to yrs of employment in industry; no measurement of ethnicity or immigrant status; considerable care, concern with methods
    Monson53     retrospective mortality cohort (SMRs) study     13,571 white male union rubber workers in any of 3 rubber plants of one company in Akron, Ohio who worked 5 yrs or more followed Jan/40-Jun/74 (5079 deaths, 37%)     age-time-cause specific mortality rate for US white males and for Ohio males     age, calendar period     no increased stomach cancer risk for all company workers; processing workers (exposure to high levels dusts, fumes including carbon black) SMR for stomach cancer = 190 (18 cases); age >74 at death SMR for GI cancer = 250     access to company and union records; concerns with methodologic issues discussed; no specific exposure measurements (but departments info available for each worker); exposures following termination unavailable for 1515 workers
    McMichael54     retrospective cohort mortality study     4 cohorts of male rubber workers in tire plants in 4 companies in Akron, Ohio followed Jan/64-Dec/74 aged 40-84 at inception (5106 deaths, 27%)     1968 race-age-cause specific US male death rates     age,race     Stomach cancer SMR ages 40-84 = 148; SMR ages 40-64 = 183, ages 65-84 = 136     calendar year not taken into account; no specific exposure measurements presented; limited analyses presented
    McMichael55     retrospective cohort mortality study     6678 male hourly rubber workers aged 40-84 Jan 1/64 at a tire plant in Akron, Ohio followed Jan/64 to Dec/72 (1783 deaths, 27%)     1968 age-cause specific US male death rates     age     Stomach cancer SMR age 40-84 = 187 (p<.001); age 40-64 SMR = 219 (p < .01)     calendar year not taken into account; no specific exposure measurements presented; limited analyses presented
    STUDIES OF MULTIPLE OCCUPATIONS
    Kneller56     case control     13,489 gastric cancer cases from the urban Shanghai cancer registry in persons over age 30 during 1980-84; census derived age-sex specific cancer incidents rates calculate for each occupational group     age-sex-site specific cancer incidence rates calculated for urban Shanghai     age, sex, occupation     Multiple statistically signif elevated SIR (standardised incidence rates) found among males: e.g. grain farmers SIR = 402; wood processing workers 176; cabinetmakers 147; sanitation workers 157; rickshaw drivers 153; leather products 150; blacksmiths 156; machine tool setters 305; sheet metal workers 140; metal grinders 141; dock workers 154; railway engine drivers and firemen 363; among females: grain farmers 414; wood processing workers 308; machinery/engine mechanics 231; sanitation workers 193; rickshaw drivers 172; plastics manufacture 160; knitters 148     case control limits; multiple comparisons; diet, smoking, home coal cooker exposures not taken into account
    Wright57     case control stomach cancer proportional incidence ratio study     1342 male cases of adenocarcinoma of the stomach in a popn based cancer registry, LA County, Calif aged 20-64 diagnosed between 1972-82 (657 cases of stomach cancer not included in study because 307 not adenocarcinomas, 350 inadequate info on occupation)     distribution of cancers in the LA County Cancer Surveillance Program     age, race, SES (based on education, income of adults in census tract of residence), immigrant status     PIR stomach cancer of exposed/PIR of unexposed = 1.3 (95% CI 1.2-1.4); for antrum/pylorus 1.8 (1.6-2.0); mineral dust 1.6 (1.0-2.5); organic dust 1.5 (1.0-2.3)     exposure to dust inferred from occupational title; occupation based only on job at time of diagnosis; men > 64 not studied because occupation not available, likely leading to underestimate of cancer and occupation link; limits of proportional incidence study; no multivariate analysis when controlling for multiple confounders
     
    MAGNESIUM PRODUCTION WORKERS
    Heldaas58     retrospective cohort cancer incidence study     2391 male Norwegian magnesium plant workers who worked more than 1 year between 1951 and 1974 followed 1953 to 1984 (exposures to volatile coal tar pitch products, chlorinated hydrocarbons (eg hexachlorbenzene), asbestos, chlorine gas, magnesium oxide and chloride dust, strong electrical magnetic fields); 21 cases of stomach cancer     national Norwegian 5 year, age specific cancer incidence rates     age, calendar year     RR for stomach cancer for overall study population = 1.6 (95% CI 1.0-2.4); stomach cancer showed highest rates in those with short and medium employment experience     Information on years of employment and age distribution of cohort not provided; methodologically sound study
    REFERENCES

    5 IS THERE EVIDENCE OF A RELATIONSHIP BETWEEN STOMACH CANCER AND OCCUPATION?

    In reviewing this literature one is struck by the huge number of studies demonstrating a statistically significant association between occupation and stomach cancer despite enormous variation in methods and approach and the inherent difficulty in demonstrating such associations in occupational cancer epidemiology. The strongest evidence comes from retrospective cohort studies in Dutch coal miners (Meijers et al), Ontario gold miners (Kusiak et al), American iron ore miners (Lawler et al), British and American rubber workers (Sorahan et al; McMichael et al; Monson et al), Norwegian magnesium production workers (Haldaas), Canadian silica exposed workers (Finkelstein et al), British foundry workers (Sorahan et al), German gas plant workers (Berger et al) and the Viennese dust exposed workers (Neuberger et al).

    These studies are supported by proportional mortality studies in American bearing plants (Park et al; Silverstein et al), female American jewellery workers (Dubrow et al), French iron ore workers (Pham), and Danish dust exposed workers (Olsen et al) and case control studies identifying increased risk of gastric cancer in Canadians exposed to wood dust (Siemiatycki et al), Californians exposed to metal dust, and British rubber workers (Coggon et al). Very large case control studies in the US and China demonstrated increased stomach cancer in those exposed to a wide range of occupational dusts (Bross et al; Kneller et al).

    While no one study is methodologically strong enough to provide unequivocal evidence of a causal relationship between stomach cancer and the occupation under study and none has quantitative exposure measurements, there are now numerous examples of excess risk of stomach cancer in various occupational groups supporting the hypothesis that occupational exposures can contribute to the development of stomach cancer.

    None of the studies was designed in such a way that one could conclude that gastric cancer was associated with a specific occupational substance identified and studied. Although Kusiak et al attempted to identify the causal agent in the gold miners they studied I believe there were significant difficulties with their underlying assumptions. They looked for substances to which gold miners were exclusively exposed compared to non-gold miners because they did not demonstrate a statistically significant gastric cancer excess in the non-gold mining groups. But analyses from this cohort have consistently shown an SMR of 130 for gastric cancer among uranium miners. Although this result is not statistically significant it may well reflect a true increase because these miners had relatively short latencies (most of the uranium mines were opened in the 1960's) and well under 20% had died at the time of the analysis. The gold mines re opened much earlier than most other mines and thus a higher proportion of dead miners included in the study with high latencies would have come from the gold mining group. This reduced the capacity of this study to demonstrate an excess of gastric cancer deaths among non-gold miners. Thus their logic was somewhat faulty and one cannot have confidence in their conclusion that chromium was the agent most strongly associated with the elevated gastric cancer rates in their gold mining cohort. Although this remains a possible explanation.

    6 EVIDENCE FOR OCCUPATIONALLY RELATED STOMACH CANCER IN CANADIAN GOLD MINERS

    As mentioned in the preface, this review is fairly limited in scope and presents an initial overview of the topic. Time constraints prevented a thorough review of the Ontario hardrock mining original data or direct communication with the authors. The following observations are based on reading the 3 reports by Muller et al, the two published papers by Kusiak et al and the IDSP report on the Ontario gold mining industry. A more thorough review of this material is warranted before compensation policy on stomach cancer is drafted. Nonetheless, the following comments are offered.

    The main evidence for stomach cancer in Canadian gold miners comes from various analyses of an Ontario cohort of hard rock miners who were followed medically by the Workers' Compensation Board. In this series of studies, jointly supported by the Ontario Ministry of Labour, the Workers' Compensation Board and the Atomic Energy Commission,59 60 61 62 63 the mortality experience of a cohort of 54,128 Ontario miners was observed. The cohort was derived from miners who were followed by the provincial chest clinic during the years 1955-1977. The first 2 reports of this study by Muller et al in 1983 and 1986 were based on mortality follow up of miners between 1955 and 1977 using somewhat different inclusion criteria. In a later update of the study the inclusion criteria for the cohort changed again and the follow up was extended to 1986. In the later analyses of this cohort published by Kusiak et al in 1991 and 1993 the cohort included miners who had attended a provincial chest clinic between 1955 and 1977, worked a minimum of two weeks in dusty jobs in non-uranium Ontario mines after 1954 and worked for a minimum of 60 months in jobs with dust exposure in non-uranium mining anywhere and/or had spent a minimum of two weeks in Ontario uranium mines after 1955 and/or had been reported exposed to radon decay products by an Ontario uranium company. In later reports (1986 and after) those who were followed by the chest clinics between 1955-1977 but no longer were mining during that period were omitted. In initial reports "gold miners" included only those who had spent 85% of their time in gold mining. This 85% rule was later removed. Non-uranium miners included those mining gold, nickel, copper, iron, lead/zinc, silver or other ores. Many miners worked in different types of mines and thus sorting out gold miners from non-gold miners is complex. The investigators do not clearly spell out how person-year exposure was attributed for those who worked in different types of mines. For example, it has been noted in a recent review of lung cancer in this cohort that person-years at risk among gold miners who later worked in uranium mines were not counted once the miner entered uranium mining64 and thus exposure years at risk among gold miners may have been underestimated.

    In the first set of reports there were 7542 defined as gold miners in the initial cohort out of a total cohort of 50,201 and they made up 15% of the cohort and contributed 17% of the person-years. In the later studies with revised definitions (Kusiak et al 1993) it appears that gold miners made up a larger proportion of the cohort and the total cohort was increased to 54,128. 9,690 deaths were observed between 1955 and 1986 in the total cohort (17.9% of the total cohort population). Although the investigators did not mention the exact number of miners in the revised gold miner cohort, the 1994 IDSP Report on lung cancer in the hardrock mining industry identifies the cohort size as 13,603 (p.18) and, thus, 25% of the total cohort. Kusiak et al in the 1993 calculate the gold miner stomach cancer SMR based on 17,207 "ever gold miners" (32% of the cohort). The total number of deaths among these miners in this analysis was not reported.

    In the last study published by this group (Kusiak et al 1993) "ever gold miners" had an overall SMR of 152 (95% C.I.125-185) for stomach cancer. The excess risk of stomach cancer was higher for those under 60 (SMR 167) than among those 60-74 years (SMR 143). Among the "ever gold miners", 6,110 had also mined uranium. This sub-group had an SMR of 171 (95% C.I. 111-252) for stomach cancer. The SMR for those born in North America was 133 and 177 for those born outside North America (compared to an SMR of 141 for non-North American born miners who had never mined gold). The excess stomach cancer risk was most marked among those with 5-19 years of latency (SMR 268 among those born in North America; SMR 280 in those born outside North America) and less evident in those with 20 or more years of latency. Non-North American born miners had a longer duration of employment underground than North American-born miners (12.4 years v.s. 9.3 years) and thus if duration of exposure were controlled for in the analysis the importance of place of birth might have diminished considerably. Calculations of stomach cancer SMR associated with varying levels of exposure showed an SMR of 211 for gold miners under 60 years of age with more than 12 person years of gold mining exposure; SMR of 144 for similarly dust exposed miners aged 60-74; SMR of 463 among those with a time-weighted index of chromium exposure of >0.3 (highest level of exposure) in gold miners less than 60 years old; SMR of 175 in similarly chromium exposed workers aged 60-74. Although no clear cut dose-response relationship was apparent, the chi-square for linear trend was highly significant for SMRs of increasing levels of chromium exposure among those under 60 years. The chi-square for linear trend was also statistically significant with increasing levels of person-years at risk reflecting dust exposure in this age sub-group. No multi-variate analyses were reported and no analyses were performed that simultaneously took into account place of birth, age, latency, calendar years exposed and exposures to dust, chromium, and uranium mining/radon.

    Overall this is a fairly sophisticated retrospective cohort study with many methodologic strengths. The cohort probably included almost all miners who met the inclusion criteria and were still working in the mines in 1954. Approximately 6% of deaths were missed. The number of miners and the total person-years worked were quite high. The cohort is believed to have included all miners who met the entry criteria because all miners were required to attend the provincial chest clinics to be certified as fit to work in jobs with dust exposure. Validation of this aspect of the Mining Master file by the IDSP showed highly accurate data65. Validation of death record certification revealed accuracy of stomach cancer diagnoses similar to that for the general Ontario male population. Age and place of birth (North America or non-North America) were controlled for in the analyses. No information on diet was available and hence was not controlled for. This study does have some weaknesses and some aspects are difficult to interpret.

    It is considerably different in design than most retrospective cohort mortality studies. Usually a workplace cohort is defined at a particular point in time and the mortality experience of all workers working at that time meeting entry criteria is followed over a designated period of time. In this cohort, new members could be added at any point between 1955 and 1977. Such members could have included new miners or previous miners who had not been working after 1954 who returned to mining sometime during 1955-1977. There is therefore the potential to underestimate an excess of stomach cancer if many new miners were added to the cohort in the later years and thus the population would tend to include many miners with relatively short latencies or duration of exposure. Given the chronology of mine closures and new mine openings this would likely affect uranium miner data most and gold miners least. In order to interpret the adequacy of these studies to answer questions about excess stomach cancer risk, it is therefore particularly important to know the number of cohort members included in the mortality study (i.e. cohort members who died) in each mining category (e.g. gold miners, uranium miners, nickel, mixed ore, etc) and the number of each of these groups in each of the latency sub-groups (e.g. number of deceased miners with latencies of < 5 years; 5-19 years; > 20-29 years; > 30 years); and the number in each of the duration of exposure sub-groups. Because exposure was considerably different over the years, sub-groups of various latency periods and duration of exposure periods must be looked at by calendar period. Such data was not supplied in the reports or published articles. It is therefore somewhat difficult to interpret all the findings or fully evaluate the methodology of this series of analyses.

    This information is particularly relevant when trying to understand the differences in stomach cancer risk among different types of miners. Crucial to the Kusiak et al hypothesis that chromium exposure accounts for the excess stomach cancer risk among gold miners is the finding that there was no excess risk among uranium miners or mixed ore miners. The capacity of this study to detect excess cancer deaths in non-gold miners may have been considerably lower than among gold miners, particularly among uranium miners who began working much later. Uranium miners had an SMR for stomach cancer of 13066 but the finding was not statistically significant. The lack of statistical significance may reflect a lower number of deceased uranium miners. Moreover if few of the uranium miners have longer latencies or duration of exposure periods, their risk may be underestimated in this study. This finding and the increased risk of stomach cancer among miners who had mined both uranium and gold suggest that excess stomach cancer risk among uranium miners cannot yet be excluded.

    In general the gold mines began operating considerably earlier than other mines (1910-1949) and closed down much earlier (1950-1970). Thus gold miners may have tended to be older than the other miners. The mines were known to have had extremely high dust levels prior to 1950 with a silicosis SMR of approx 850, high rates of silicotuberculosis and measured dust levels in the 1930's and 40's often above 1000 particles/ml. Although dust levels in gold mines and non-gold mines may have been comparable at any given period, dust exposure levels were likely much higher among miners who worked prior to 1950 and thus the cumulative exposure levels were likely highest among gold miners included in the mortality study.

    Although these investigators did have access to some quantitative exposure measurements by occupation for dust and radon, they chose not to make use of it: "The results of the recent studies on nickel miners (Ro82, Sh82) did not indicate that a mortality study aimed at establishing the quantitative relationship between exposure to various environmental factors and death from occupational disease would serve a useful purpose...No assessment, in quantitative terms, was, therefore, made of the exposure of non-uranium miners to various dusts in the mine environment."67 Instead they used years of mining experience (person-years at risk, PYR) as a proxy measure of dust exposure. Thus information about variations in exposure over time (eg it is known that for some jobs there was more than a 5-fold difference in dust levels in the 1940's compared to 1967)68, among mining camps, among types of mines by ore mined, and among occupations within individual mines was not taken into account. Time-weighted indices were developed for arsenic and chromium exposure in analyses published by Kusiak et al in 1991 and 1993. The lack of accurate exposure measurements in this study is its most significant weakness. The lack of accurate quantitative exposure estimates for each cohort member makes it difficult to evaluate dose-response relationships between exposures and outcomes and makes it very difficult to sort out which of the many specific exposures is associated with the excess stomach and lung cancer deaths observed.

    The following speculations are offered with the proviso that their relevance can only be judged following more detailed knowledge of the original data. The gold miners were older and had had longer periods of exposure than the other miners. They began gold mining many decades earlier than other miners. But because registry information was only available from the 1950's the miners with heavy exposure in the early decades of the industry who were no longer working in the mines after 1954 were not included. While this decision was reasonable, it may have eliminated those in the registry with the highest dust exposures and might have contributed to the very surprising result of a lower SMR for stomach cancer among those in the study who began working prior to 1945. This may have influenced the results and the difficulty finding a dose response relationship. Alternative explanations are also possible, including aluminum powder exposure which began around 1945. This latter hypothesis was ruled out by Kusiak et al but the relevant analysis was not described.

    Despite these limitations this study consistently demonstrates a significant increase in stomach cancer in gold miners. The finding of a greater excess of stomach cancer in younger workers and those with 5 to 20 years latency is surprising and difficult to interpret. It may reflect a group of workers with very high exposure and suggest that intensity of exposure may be more relevant than duration of exposure. Or it may reflect unmeasured interactions (eg if combined exposure to gold and uranium mining increases risk considerably, and mainly younger gold workers were hired by uranium mines after gold mine closures, then we might see the results reported by Kusiak et al.). There may well be a need for more complex analyses that simultaneously take into account a wider range of exposure variables and risk modifiers.

    Although not reviewed in section 4 because the studies were not obtained in time for the initial review, there are four other international studies of cancer mortality among gold miners, two of which have shown an increased risk of stomach cancer69 70 71 72. Methodologic limitations or significant differences in exposure make the generalizability of these studies to the Ontario situation limited. Nonetheless they do indicate that excess risk of stomach cancer in gold miners has been noted previously elsewhere.

    7 IS THERE EVIDENCE OF A CAUSAL RELATIONSHIP BETWEEN STOMACH CANCER AND OCCUPATIONAL EXPOSURE IN ONTARIO GOLD MINERS?

    The following comments on causation based on the Muller-Kusiak et al studies are offered with full acknowledgement that a more knowledgeable review of this question can be made only with more detailed review of the original data upon which the reports were based. These comments should be seen as preliminary.

    In order to demonstrate a causal association between stomach cancer and working in Ontario gold mines the following criteria elaborated by Sackett et al73 and adapted from the Bradford Hill criteria for causation74 are useful:

    Design: For ethical reasons there can, of course, be no studies based on human experiments to address this question. Thus study design for this question is limited to cohort and case control studies. A retrospective cohort study design is highly appropriate and likely to provide the strongest evidence feasible.

    Strength of association: Kusiak et al have demonstrated an overall SMR for gastric cancer among Ontario gold miners of 152 (95% CI 125-185). Sub-groups by age and exposure to chromium or dust have somewhat higher SMRs. The highest SMR, 463 (no confidence levels stated) was seen among those under 60 years of age and exposed to the highest estimated levels of chromium. The strength of association in this study is generally moderately high demonstrating a 50 to 70% increase in stomach cancer risk. None of the SMRs reported simultaneously took into account age, immigrant status, latency, nature and duration of exposures. This would be useful in order to have a more accurate idea of excess risk.

    Consistency among studies: There are four other studies of stomach cancer and gold miners. All have serious methodologic limitations or deal with exposures not relevant to the Ontario context (eg asbestos). A study of South African gold miners showed an excess risk of stomach cancers which was not statistically significant. A Soviet study showed a very high and statistically significant relative risk for stomach cancer in gold miners. While there are numerous studies in Table 1 of other occupations in which dust exposure was prominent and an excess of stomach cancer risk was demonstrated, negative studies were also found. These tended in general to be of slightly lower quality but no systematic ranking of quality was performed.

    Biological plausibility: An association between occupation and stomach cancer makes both epidemiologic sense and is biologically plausible. The stomach is one of the organs which can have significant exposure from occupational hazards through inhalation and subsequent swallowing of material that is coughed up or enters the upper respiratory tract and through ingestion of contaminants that are on workers' hands when they eat or smoke. The various theories of occupational carcinogenesis proposed by Ames and others mentioned in Section 3 above demonstrate that occupational gastric cancer is quite biologically plausible. As mentioned above and detailed in Table 1 there have been many consistent associations found between workers in occupations exposed to dusts, hydrocarbons and/or nitrosamines or their precursors in working populations in many different countries by research groups employing different methods to carry out their studies.

    Specificity: As mentioned in the introduction, gastric cancer likely is multifactorial in etiology and thus there is no one specific cause.

    Temporal relationship: Kusiak et al are able to show a somewhat appropriate temporal relationship between occupational exposure and subsequent development of gastric cancer. Latencies of 5-19 years were associated with higher risk than latencies of greater than 20 years.

    Dose-response relationship: The greatest weakness in the Muller -Kusiak et al studies is the failure to demonstrate a dose-response relationship. This may well be due to the lack of adequate exposure measurement. It may also reflect the multi-factorial nature of stomach cancer and the interactions among several contributing factors.

    Analogy: The findings of excess cancer risk among gold miners is coherent with other studies which have demonstrated similar risk among There is considerable coherence and consistency among findings and none of the evidence conflicts with other accepted facts about stomach cancer epidemiology and pathogenesis.

    8 CONCLUSIONS

    Given the limited nature of this review, the following conclusions must be seen as preliminary or tentative at best.

    1 There is consistent evidence of a relationship between stomach cancer and occupation and existing literature supports the potential work-relatedness of stomach cancer. The study by Kusiak et al75 provide strong evidence for an excess of stomach cancer among Ontario gold miners. Although the findings do not provide a clear dose-response relationship, revised analyses using more appropriate entry criteria for the gold miner cohort and extending the follow up period have strengthened the observations of excess stomach cancer risk. Alternative explanations for these observations are not readily apparent. Nutritional hypotheses have not been explored. The investigators have demonstrated that immigrant status alone cannot account for the excess risk. Ontario gold miners have been exposed to several important known and suspected carcinogens including silica and other dusts, radon decay products, chromium, arsenic, asbestiform fibres, and PAHs. Two or more of these agents may interact and miners may have been differentially exposed to these agents at different time periods or different mines, thus possibly accounting for the surprising results with respect to latency and duration of exposure.

    2 Many of the criteria for causality have been met but not all - in particular a clear dose response relationship has not yet been demonstrated. While an unequivocal causal relationship between an occupational exposure and stomach cancer has not been demonstrated, the consistency of the findings make it unwarranted to dismiss this probable connection. The lack of a clear dose-response relationship may reflect the lack of adequate exposure measurement in studies conducted to date rather than the non-existence of such a relationship. Dose-response relationships in this population may be better evaluated with revised analyses using improved estimates of exposure.

    3 The epidemiological studies of occupational stomach cancer have not been designed in such a way as to be able to implicate any specific occupational hazardous substance. The studies are consistent with hypotheses previously proposed in the literature by Ames including poly-aromatic hydrocarbons; nitrosamines; and freshly generated dust. Chromium exposure remains a possible explanation of excess gastric cancer in Ontario gold miners but alternative occupational exposures are also possible explanations. Uranium mining exposures and radon in gold mines may also play a contributing role. Other possibilities exist such as the interaction among occupational dust (or specific hazardous components of dust), nutritional exposures and previous infection with helicobacter pylori.

    9 RECOMMENDATIONS

    1. The literature reviewed above does not provide a definitive answer about the nature of the relationship between gastric cancer and occupational exposure in gold or uranium miners in Ontario. But it does suggest that such a relationship may well exist and cannot be ruled out. In order to address the issue of compensability of stomach cancer among these miners, it may be prudent to re-analyze the Muller, Kusiak et al data using quantitative exposure estimates. A skilled industrial hygienist trained in risk assessment modelling may well be able to develop validated exposure estimates based on existing hygiene measurements, reports documenting conditions at the time and interviews with surviving miners and others from the mining industry. Such estimates could probably be derived for dust, radiation, chromium, arsenic and other exposures deemed appropriate. Consultation with a senior biostatistician experienced in cancer mortality studies may allow for multivariate analyses that simultaneously take into account age, latency, duration of exposure, and calendar period exposed, immigrant status, and various exposure measures. It may be useful to explore whether nutritional patterns among miners (eg the main foods available to the miners) was substantially different from nutritional patterns in the rest of the province during the relevant time periods.

    2. As mentioned previously, this review is only a partial review of the literature. If the Panel is interested in addressing the question of stomach cancer among a wider range of occupations, a more systematic review of the literature will be needed with formal evaluation of the methodologic quality of each relevant study. Such a review should address each exposure or occupation of interest separately.

    3. In carrying out this review as requested, the comprehensiveness of the review and its relevance for the Panel's deliberations on stomach cancer in gold miners is not as high as it could have been had a more limited question been negotiable or adequate consulting time allowed for rating of each article and a more systematic synthesis of the results. The standards for literature reviews have changed considerably in the past 5 to 10 years. Current recommended methods include the identification of a precise question, the search for all potentially relevant studies (published and unpublished), a selection of the most relevant papers on the basis of clearly identified criteria, systematic evaluation of these studies for methodologic quality, and synthesis and drawing of conclusions based on the highest quality studies. It appeared that the IDSP policy staff with whom I had contact were unfamiliar with this approach. The necessity of acquiring each of the articles identified as potentially relevant in the bibliographic search was not appreciated at all. Moreover the 1987 IDSP report on this topic was not sent nor was its existence mentioned even though it was unlikely to appear in any computer-based bibliographic search. I happened on it fortuitously but somewhat later than would have been optimal. The ethics of not passing on this published report particularly as it would have hastened understanding of the context in which the review was requested is questionable and demonstrates a lack of appreciation of the need for comprehensiveness in a literature review. Perhaps a presentation by an epidemiologist who teaches systematic overview techniques could help IDSP staff members better formulate the questions they request for review and better appreciate the need for comprehensive searches for relevant papers in order to decreases the potential for bias in selection of papers for review. Dr. Andy Oxman at McMaster University would likely be the ideal candidate to provide such training to IDSP staff.

    10 APPENDIX A  RELEVANT REFERENCES IDENTIFIED BUT NOT REVIEWED

    Armstrong BK, McNulty JC, Levitt LJ, Williams KA, Hobbs MS. Mortality in gold and coal miners in Western Australia with special reference to lung cancer. Br J Indust Med 1979, 36, 199-205

    Ashley BJB. Environmental factors in the etiology of gastric cancer. Br J Prev Soc Med 1969, 23, 187-9

    Atuhaire LK, Campbell MJ, Cochrane AL, et al. Mortality of men in the Rhondda Fach 1950-80. Br J Indust Med 1985, 42, 741-5

    Blair A. Mortality among workers in the metal polishing and plating industry 1951-69. J Occup Med 1980, 22, 158-162

    Bridbord K, Costello J, Gamble, et al. Occupational safety and health implications of increased coal utilization. Environ Health Perspectives 1979, 33, 285-302

    Brown DP, Kaplan SD, Zumwalde RD, Archer VE. Retrospective cohort mortality study of underground gold mine workers. In Proceedings of the NCI/EPA/NIOSH Collaborative Workshop: Progress on Joint Environmental and Occupational Cancer Studies, pp 7-55

    Choi N. Ethnic distribution of cancer of the gastrointestinal tract in Manitoba. Am J Public Health 1968, 58, 2067-80

    Clemmesen J, Nielson A. The social distribution of cancer in Copenhagen 1943 to 1947. Br J Cancer 1951, 5, 159-171

    Costa D, Guignard J, Zalma R, Pezerat H. Production of free-radicals arising from surface activity of minerals toward oxygen. Part I: case of iron mine ores. Toxicol Ind Health 1989, 5

    Craven JL, Baum M, West RR. Variations in gastric cancer incidence in South Wales. Clin Oncol 1979, 5, 3431-51

    Creagan BT, Hoover RN, Fraumeni JF Jr. Mortality from stomach cancer in coal mining regions. Arch Environ Health 1974, 28, 28-30

    Decoufle P. Further analysis of cancer mortality patterns among workers exposed to cutting oil mists. JNCI 1978, 61, 1025-1030

    Delzell IS, Monson RR. Mortality among rubber workers: V Processing workers. J Occup Med 1982, 24, 539-545

    Dubrow R, Wegman DH. Cancer and occupation in Massachusetts: a death certificate study. Am J Indust Med 1984, 6, 207-230

    Dubrow R, Wegman D. Setting priorities for occupational cancer research and control-synthesis of the results of occupational disease surveillance studies. JNCI 1983, 71, 1123-42

    Enterline PE. Mortality rates among coal miners. Am J Public Health 1964, 54, 758-68

    Finkelstein MM, Kusiak RA, Suranyi G. Mortality among miners receiving workers' compensation for silicosis in Ontario: 1940-75. J Occup Med 1982, 24, 663-7

    Fox AJ, Goldblatt P, Kinlen IJ. A study of the mortality of Cornish tin miners. Br J Indust Med 1981, 38, 378-380

    Hanis NM, Stavraky MK, Fowler JL. Cancer mortality in oil refinery workers. J Occup Med 1979, 21, 167-174

    Hass JF, Schottenfeld D. Epidemiology of gastric cancer. In Lipkin M, Good RA (eds): Gastrointestinal Tract Cancer, Plenum, New York, 1978

    Hayes RB, Sheffet A, Spiritas R. Cancer mortality among a cohort of chromium pigment workers. Am J Ind Med 1989, 16, 127-133

    Houten L, Bross IDJ, Viadana E, et al. Occupational cancer in men exposed to metals. Adv Exp Med Biol 1977, 91, 93-102

    Jacobsen M. Dust exposure, lung diseases and coal miners mortality. Edinburough: Univ of Edinborough, 1976, PhD thessis

    Jarvholm B, Lillienberg L, Stallsten G, Thiringer G, Axelson O. Cancer morbidity among men exposed to mist oil in the metal industry. J Occup Med 1981, 23, 333-337

    Jarvholm B, Thiringer G, Axelson O. Cancer morbidity among polishers. Br J INdusst Med 1982, 39, 196-7

    Katsnelson BA, Mokronosova KA. Non-fibrous mineral dust and malignant tumors. J Occup Med 1979, 21, 15-20

    Kinlen IJ, Willows AN. Decline in the lung cancer hazard: a prospective study of the mortality of iron ore miners in Cambria. Br J Ind Med 1988, 45, 219-24

    Klauber MR, Lyon JL. Gastric cancer in a coal minig region. Cancer 1978, 41, 2355-8

    Koselka RS, Hernberg S, Karava R, et al. A mortality study of foundry workers. Scand J Work Environ Health 1976, 2, suppl 1, 73-89

    Kusiak RA, Springer J, Ritchie AC, Muller J. Lung cancer in Ont gold miners: possible etiologic factors. Br J Ind Med 1991, 48, 808-17

    Lerer T, Redmond C, Cambell J. Factors related to the incidence of stomach cancer in the Pittsburgh Standard Metropolitan Statistical Area. J Natl Cancer Inst 1977, 59, 1065-1971

    Lindsay JP, Stavraky KM, Howe GR. The Canadian labour force 10% Sample Study: Cancer mortality among men, 1965-1979. J Occup Med 1993, 35, 408-414

    Mancuso TF, Ciocco A, El-Attar AA. An epidemiologic approach to the rubber industry. J Occup Med 1968, 10, 213-232

    Matalo NM, Klauber MR, Gorishek WM. High incidence gastric carcinoma in a coal minig region. Cancer 1972, 29, 733-737

    McLaughlin AG, Harding HE. Pneumoconiosis and other causes of death in iron and steel foundry workers. Arch Ind Health 1956, 14, 350-378

    McMichael AJ, Spiritas R, Kupper LL, Gamble JF. Solvent exposure and leukemia among rubber workers. J Occup Med 1975, 17, 234-239

    Milham S. A study of AFL-CIO United Brotherhood of Carpenters and Joiners of America, 1969-1970. DHEW publication no.(NIOSH) 74-129, Springfield VA, National Technical Info Service 1974

    Muller J, Kusiak RA, Suranyi G, Ritchie AC. Study of mortality of Ont miners 1955-77, Addendum. Toronto, Ont Ministry of Labour, Ont WCB, Atomic Energy Control Bd 1987

    Mur J, Meyer-Busch C, Pham QT et al. Risk of lung cancer among iron ore miners: a proportional mortality study of 1075 deceased miners in Lorraine, France. J Occup Med 1987, 29, 762-768

    Newman AM, Spengler RF. Cancer mortality among immigrant populations in Ontario, 1969-73. Can Med Assoc J 1984, 130, 399-405

    Pippard EC, Acheson ED, Winter PD. Mortality of tanners. Br J Indust Med 1985, 42, 285-7

    Registrar General's Decennial Supplement, England and Wales 1961 (1971): Occupational mortality tables. Her Magesty's Stationery Office, London, England

    Schneiderman MA. Digestive sytem cancer among persons subjected to occupational inhalation of asbestos particles: a literature review with emphasis to dose response. Environ Health Perspect 1974, 9, 307-315

    Selikoff IJ. Epidemiology of gastrointestinal cancer. Environ Health Perspect 1974, 9, 299-306

    Simanato L, Ferro G, Winkelman R, Saracci R. A mortality study of miners and factory workers at the "societe des mines et produit chemiques de Salsigne", France. Lyon, France, IARC, WHO

    Sparks PJ, Wegman DH. Cause of death among jewellery workers. J Occup Med 1980, 22, 733-766

    Spiegelhalder B, Muller J, Drasche H, Preussmann R. N-nitrosodiethanolamine excretion in metal grinders. In Bartsch H, O'Neill I, Schulte-Hermann R (eds): The Relevance of N-Nitroso Compounds to Human Cancer. IARC Sci Pub No. 84, 1987, 550-552

    Stellman SD, Garfinkel L. Cancer mortality among woodworkers. Am J Indus Med 1984, 5, 343-57

    Stewart HL. Experimental alimentary tract cancer. NCI Monogr 1967, 25, 199-217

    Thomas TL, Waxweiler RJ, Crandell MS et al. Cancer mortality patterns among workers in 3 Texas oil refineries. Am J Indust Med 1984, 6, 3-16

    Vena JE, Sultz HA, Fiedler RC, et al. Mortality of workers exposed to an anti-trust oil. Scand J Work Environ Health 1981, 7, 179-184

    Waldron HA. Health car of people at work: exposure to oil mist in industry. J Soc Occup Med 1977, 27, 45-49

    Walrath J, Rogot E, Murray J, Blair A. Mortality patterns among US veterans by occupation and smoking status. NIH Publication no. 85-2756, 1985, 38

    Wang JD, Wegman DH, Smith T. Cancer risks in the optical manufacturing industry. Br J Ind Med 1983, 40, 177-181

    Wegman DH, Eisen E. Causes of death among employees of a synthetic abrasive product manufacturing company. J Occup Med 1981, 23, 748-754

    Winkelstein W, Kantor S. Stomach cancer - positive association with suspended particulate air pollution. Arch Environ Health 1969, 18, 544-547

    Wooley PV, Pinsky SD. Binding of N-nitroso carcinogens in pancreatic tissue. Cancer 1981, 47, 1485-9

    Wyndham CH, Bezuidenhout BN, Greenacre MJ, Sluis-Cremer GK. Mortality of middle aged white South African gold miners. Br J Indust Med 1986, 43, 677-684

    REFERENCES

    1.Kusiak RA, Springer J, Ritchie AC, Muller J. Carcinoma of the lung in Ontario gold miners: possible aetiologic factors. Br J Indust Medicine 1991, 48, 808-817

    2.Parkin D, Laea E, Muir C. Estimates of the worldwide frequency of 16 major cancers in 1980. Int J Cancer 1988, 41, 184-97

    3.Page HS, Asire AJ. Cancer rates and risks, Third edition. 1985. NIH publication no. 85-691, US Department of Health and Human Services

    4.Nutt A. Rubber work and cancer - past, present and perspectives. Scand J Work Environ Health 1983, 9, suppl 2, 49-57

    5.Kraus AS, Levin ML, Gerhardt PR. A study of occupational associations with gastric cancer. Am J Public Health 1957, 47, 961

    6.Hagstrom RM, Sprague HA, Landau E. The Nashville air pollution study. VII. Mortality from cancer in relation to air pollution. Arch Environ Health 1967, 15, 237

    7.Winkelstein W Jr, Kantor S. Stomach cancer. Positive association with suspended particulate air pollution. Arch Environ Health 1969, 18 544

    8.Armijo R, Gonzales A, Orlellana M et al. Epidemiology of gastric cancer in Chile: II - nitrate exposures and stomach cancer frequency. Intl J Epidemiology 1981, 10, 57

    9.Meyer MB, Luk GD, Sotelo JM, et al. Hypothesis: the role of the lung in stomach carcinogenesis. Am Rev Respir Dis 1980, 121, 887-892

    10.Gamble JF, Ames RG. The role of the lung in stomach carcinogenesis: a revision of the Meyer hypothesis. Medical Hypotheses 1983, 11, 359-364

    11.Ames RG, Gamble JF. Lung cancer, stomach cancer, and smoking status among coal miners: a preliminary test of a hypothesis. Scand J Work, Environ, Health 1983, 9,443-48

    12.Kusiak et al 1993 Ibid

    13.Neugut AI, Wylie P. Occupational cancers of the gastrointestinal tract. Occupational Medicine: State of the Art Reviews 1987, 2, 109-135

    14.Nutt A. Rubber work and cancer - past, present and perspectives. Cancer Mortality and Morbidity 1983, 9, 49-57

    15.IDSP. Report to the Workers' Compensation Board on the Ontario Gold Mining Industry, IDSP Report No 1, Toronto, Ont, April 1987

    16.Kusiak RA, Ritchie AC, Springer J, Muller J. Mortality from stomach cancer in Ontario miners. Br J Indust Med 1993, 50, 117-126

    17.Kusiak RA, Springer J, Ritchie AC, Muller J. Carcinoma of the lung in Ontario gold miners: possible aetiological factors. Br J Indus Med 1991, 48, 808-817

    18.Muller J, Kusiak RA, Suranyi G, Ritchie AC. Study of Mortality of Ontario Gold Miners 1955-1977. Feb, 1987, Ontario Ministry of Labour, Workers' Compensation Board of Ontario, Atomic Energy Control Board of Canada

    19.IDSP. Report to the Workers' Compensation Board on the Ontario Gold Mining Industry. IDSP Report No 1, Toronto, Ont, Apr, 1987

    20.Muller J, Wheeler WC, Gentleman JF, Suranyi G, Kusiak RA. Study of Mortality of Ontario Miners 1955-1977. Part I. April, 1983, Ontario Ministry of Labour, Workers' Compensation Board of Ontario, Atomic Energy Control Board of Canada

    21.Hodgeson JT, Jones RD. Mortality of a cohort of tin miners 1941-86. Br J Indust Med 1990, 47,665-676

    22.Lawler AB, Mandel J, Schuman LM, Lubin JH. A retrospective cohort mortality study of iron ore (hematite) miners in Minnesota. J Occup Med 1985, 27, 507-517

    23.Atuhaire LK, Campbell MJ, Cochrane AL et al. Gastric cancer in a south Wales valley. Br J Indust Med 1986, 43, 350-52

    24.Swaen GMH, Aerdts WHM, Sturmans F, et al. Gastric cancer in coal miners: a case control study in a coal mining area. Br J Indus Med 1985, 42, 627-630

    25.Swaen GMH, Aerdts, Slangen JJM. Gastric cancer in coal miners: final report. Br J Indust Med 1987, 44, 777-779

    26.Meijers JMM, Swaen GMH, Slangen JJM, et al. Long-term mortality in miners with coal workers' pneumoconiosis in the Netherlands: a pilot study. Am J Indust Med 1991, 19, 43-50

    27.Pham QT, Chau N, Patris A et al. Prospective mortality study among iron miners. Cancer Detection and Prevention 1991, 15, 449-454

    28.Weinberg GB, Kuller LH, Stehr PA. A case control study of stomach cancer in a coal mining region of Pennsylvania. Cancer 1985, 56, 703-713

    29.Berger J, Manz A. Cancer of the stomach and colon-rectum among workers in a coke gas plant. Am J Indust Med 1992, 22, 825-834

    30.Siemiatycki J, Richardson L, Gerin M, et al. Associations between several sites of cancer and nine organic dusts: results from an hypothesis-generating case control study in Montreal, 1979-83. Am J Epidemiol 1986, 123,235-49

    31.Milham S, Demers RY. Mortality among pulp and paper workers. J Occup Med 1984, 844-6

    32.Roscoe RJ, Steenland K, McCammon CS, et al. Colon and stomach cancer mortality among automotive wood model makers. J Occup Med 1992, 34, 759-770

    33.Svirchev LM, Gallagher RP, Band PR et al Gastric Cancer and lymphosarcoma among wood and pulp workers. J Occup Med 1986, 28, 264-5

    34.Olsen JH, Moller H, Jensen OM. Risks for respiratory and gastric cancer in wood-working occupations in Denmark, J Cancer Res Clin Oncol 1988, 114,420-424

    35.Toren K, Sallsten G, Jarvholm B. Mortality from asthma, COPD, respiratory system cancer, and stomach cancer among paper mill workers: a case-referent study. Am J Indust Med 1991, 19, 729-737

    36.Finkelstein M, Liss GM, Krammer F, Kusiak RA. Mortality among workers receiving compensation award for silicosis in Ontario 1940-85, Br J Indust Med 1987, 44, 588-594

    37.Coggon D, Barker DJP, Cole RB. Stomach cancer and work in dusty industries. Br J Indust Med 1990, 47, 298-301

    38.Neuberger M, Kundi M. Occupational dust exposure and cancer mortality - results of a prospective cohort study. In Fletcher AC, Saracci R, Thomas TL (ed): Occupational Exposure to Silica and Cancer Risk. Lyon, International Agency on Cancer, 1990, 65-73

    39.Neuberger M, Kundi M, Westphal G, Grundorfer W. The Viennese Dusty Worker Study. in: Silica, Silicosis and Cancer, 1986, 415-422

    40.Bross ID, Viadana E, Houten L. Occupational cancer in men exposed to dust and other environmental hazards. Arch Environ Health 1978, 33,300-307

    41.Sorahan T, Cooke MA. Cancer mortality in a cohort of United Kingdom steel foundry workers; 1946-85. Br J Indust Med 1989, 46, 74-81

    42.McDowell ME. A mortality study of cement workers. Br J Indust Med 1984, 41, 179-182

    43.Amandus HE. Mortality from stomach cancer in US cement plant and quarry workers, 1950-80. Br J Indust Med 1986, 43, 526-528

    44.Wu-Williams AH, Yu MC, Mack TM. Lifestyle, workplace and stomach cancer by subsite in young men of Los Angeles County. Cancer Research 1990, 50, 2569-76

    45.Dubrow R, Gute DM. Cause-specific mortality among Rhode Island jewellery workers. Am J Indust Med 1987, 12, 579-593

    46.Sparks PJ, Wegman DH. Cause of death among jewelry workers. J Occup Med 1980, 22, 733-736

    47.Garabrant DH, Wegman DH. Cancer mortality among shoe and leather workers in Massachusetts. Am J Indust Med 1984, 5, 303-314

    48.Edling C, Kling H, Axelson O. Cancer mortality among leather tanners. Br J Indust Med 1986. 43, 494-496

    49.Park RM, Wegman DH, Silverstein MA, et al. Causes of death among workers in a bearing manufacturing plant. Am J Indust Med 1988, 13, 569-680

    50.Silverstein M, Park R, Marmor M et al. Mortality among bearing plant workers exposed to metalworking fluids and abrasives. J Occup Med 1988, 30, 706-714

    51.Sorahan T, Parkes HG, Veys CA, et al. Mortality in the British rubber industry 1946-85. Br J Indust Med 1989, 46, 1-11

    52.Sorahan T, Parkes HG, Veys CA, Waterhouse JAH. Cancer mortality in the British rubber industry 1946-80. Br J Indust Med 1986, 43, 363-373

    53.Monson RR, Nakano KK. Mortality among rubber workers: 1. White male union employees in Akron, Ohio. Am J Epidemiology 1976, 103, 284-296

    54.McMichael AJ, Andjelkovic DA, Tyroler HA. Cancer mortality among rubber workers: an epidemiologic study. Annals New York Academy of Sciences 1976, 125-137

    55.McMichael AJ, Spirtas R, Kupper LL. An epidemiologic study of mortality within a cohort of rubber workers, 1964-72, J Occup Med 1974, 16, 458-464

    56.Kneller RW, Gao YT, McLaughlin JK, et al. Occupational risk factors for gastric cancer in Shanghai, China. Am J Indus Med 1990, 18, 69-78

    57.Wright WE, Bernstein L, Peters JM. Adenocarcinoma of the stomach and exposure to occupational dust. Am J Epidemiology 1988, 128, 64-73

    58.Heldaas SS, Langard S, Andersen. Incidence of cancer in a cohort of magnesium production workers. Br J Indust Med 1989, 46, 617-623

    59.Muller et al 1983 Ibid

    60.Muller et al, 1986 Ibid

    61.Muller et al, 1987, Ibid

    62.Kusiak et al, 1991, Ibid

    63.Kusiak et al, 1993, Ibid

    64.IDSP. Report to the Workers' Compensation Board on Lung Cancer in the Hardrock Mining Industry, March 1994, IDSP Report No. 12, p. 19

    65.IDSP Report no 1, 1987 Ibid

    66.Muller et al 1983 Ibid, p. 75 (Table 14)

    67.Muller et al 1983 Ibid, p. 12

    68.Kusiak et al 1991 Ibid

    69.Armstrong BK, McNulty JC, Levitt LJ, Williams KA, Hobbs MS. Mortality in gold and coal miners in Western Australia with special reference to lung cancer. Br J Indust Med 1979, 36, 199-205

    70.Simanato L, Ferro G, Winkelman R, Saracci R. A mortality study of miners and factory workers at the "societe des mines et produit chemiques de Salsigne", France. Lyon, France, IARC, WHO

    71.Katsnelson BA, Mokronosova KA. Non-fibrous mineral dust and malignant tumors. J Occup Med 1979, 21, 15-20

    72.Brown DP, Kaplan SD, Zumwalde RD, Archer VE. Retrospective cohort mortality study of underground gold mine workers. In Proceedings of the NCI/EPA/NIOSH Collaborative Workshop: Progress on Joint Environmental and Occupational Cancer Studies, pp 7-55

    73.Sackett DL, Haynes RB, Tugwell P. Clinical Epidemiology: a Basic Science for Clinical Medicine, Little, Brown and Co., Toronto, 1985

    74.Hill AB. Principles of Medical Statistics (9th edition). Lancet, London 1971, p. 313

    75.Kusiak et al 1993