MATERIALS AND METHODS

The base population for this historical cohort comprised all Swedish men and women who were gainfully employed at the time of the 1970 census, had also been recorded in the 1960 census and were still alive and in the age range of 25–59 on 1 January 1971. A historical cohort was formed by all Swedish people by occupation according to the census of 1970, 1 779 646 men and 1 066 346 women, and inside it a subcohort including only those referring the same occupation in both the 1960 and 1970 censuses (comprising 693 501 men and 231 858). The occupational cohorts were followed up from 1971 to the end of 1989. Death was the only cause of censoring considered. People not reported as dead during the follow up were considered to be alive at the end of follow up.

Information was drawn from two data sets. Firstly, the Swedish cancer environment register, providing information on incident cancer cases, occupation, and demographic variables such as age and residency from the 1960 and 1970 censuses; this register was used to compute specific rate numerators, with gastric cancer taken to be any case with code 151 of the seventh revision of the international classification of diseases (ICD-7). The Swedish cancer registry provides valuable information on cancer incidence, with high reporting rates: 95%–98% of all tumours are estimated to be reported.^35,36 Secondly, a background population register, comprising all people in the 1970 census, with information on occupation and residence in 1970, occupation in 1960, and date of death; this register was used to compute specific rate denominators.

In the 1970 census, occupations were coded according to the Nordic classification of occupations.³⁷ Similar codes were used in the 1960 census and translations were made to the 1970 code when necessary. Every occupation is represented by a three digit number. The first digit refers to one of 10 major occupational sectors (0–9), where higher numbers indicate manual occupations, and lower numbers non-manual occupations often requiring longer education associated with a higher socioeconomic status.

The exact number of person-years was calculated with Clayton's algorithm³⁸ by sex, age group (25–29, 30–34, ..., 75–79), occupation coded as in the Nordic classification of occupations³⁷ on a three digit basis, calendar period (1971–5, 1976–80, 1981–5 y 1986–9) and county for the general cohort as well as the subgroup of people reporting the same occupation in both censuses, henceforth called the subcohort.

On the assumption that the number of cases in each age group, sex, and calendar period would be distributed as a Poisson variable, log linear Poisson models were fitted in both sexes and cohorts for the calculation of occupational relative risks (RRs), comparing occupations against others in the cohort. In this model, the number of expected cases was introduced as an offset.³⁸ Given that the expected number was computed on the basis of the age and period specific reference rates, the RR for each occupation was likewise adjusted for age and period. The RRs were estimated for occupational codes by one, two, and three digit (occupational sector, occupational group, and occupation, respectively) taking all other occupations as the reference rate.

In view of the fact that gastric cancer risk shows a geographical variation, the RRs were adjusted for geographical risk area. As occupations are not uniformly distributed geographically, we decided to group counties according to geographical risk areas (based on their standardised incidence ratio (SIR)). The SIR is the ratio of the observed to the expected number of cases in a specific area. Sweden is divided into 24 counties that were consequently grouped into five categories depending on the observed SIR value: (a) less than 0.85, (b) between 0.85 and 0.95, (c) between 0.95 and 1.05, (d) between 1.05 and 1.15, and (e) more than 1.15. These five categories were represented by dummy variables in the regression.

Finally, as a means of assessing the influence from factors associated with social class—for example, dietary habits—the same analysis was further run with other occupations in the same sector as the reference rate (namely, the corresponding one digit category according to the Nordic classification of occupations, except for production sectors I and II that were considered together). However, people in the same occupational sector may also share relevant occupational exposures and in this case the second analysis would underestimate their risk. For this reason both RRs are presented. Because of the many occupational categories, only those with 1000 or more people and at least 10 cases of gastric cancer in the general cohort were evaluated.

RESULTS

There were 8763 cases of gastric cancer among men and 2725 among women in the study population. Significant excesses of stomach cancer were found for men in the occupational sectors (one digit) mining and quarrying and production (table 1), the overall cohort was used to provide reference rates. For women, a significant excess was found among service workers.

Men

Table 2 shows the RR for Swedish men by occupational groups (two digit). People in professional and technical, administrative and managerial, bookkeeping, clerical, and the sale sectors generally showed reductions in risk of gastric cancer compared with the whole cohort combined. In the case of engineering workers, when taking as reference those occupations in the same sector (those with the same first digit) engineering work registered a significant excess risk of 25% for people in the general cohort and of 36% in those reporting such occupations in 1960 and 1970. By contrast, the incidence was markedly decreased among medical workers even when only other occupations in the same main group were taken as reference.

For agricultural, forestry, and fishing sector, fishing work showed an excess risk in both the general cohort and those in the same occupations in 1960 and 1970, with similar increases of more than 30% when people of the same occupational sectors were used as reference. Miners and quarrymen employed for 10 or more years in this occupational group ran the highest risk of gastric cancer (RR 1.91). For this group, it is not possible to calculate RRs within the sector, given that it is the only occupational group in this sector. Within the transport and communication sector, the title rail and road transport work showed significant excess in both cohorts, whereas ships' crew presented non-significant increased risk in all analyses.

For the the production sector, significant excess risks were found in the general cohort for masonry and concrete work, heavy labour, and warehouse and packaging work. We also found non-significantly increased risks for textile work, sewing, shoe and leather goods work, metal processing work, and glass, ceramic, and food industries when the overall cohort provided the reference rates, although risks were attenuated when people of the same occupational sector were used as reference. The analysis of all people who reported the same occupation in 1960 and 1970 gave a more specific selection of occupational groups with high risk, this time textile, metal processing, masonry and concrete, glass or ceramics, and warehouse or packaging work stood out. Among service occupations, incidence of gastric cancer was increased for housekeeping workers and waiters.

Table 3 examines three digit occupations for those codes with an RR greater than 1.10 in men. Significant increases in gastric cancer risk appeared in the general cohort among fishermen (RR 1.41), miners and quarrymen (RR 1.78), motor vehicle drivers (RR 1.13), sheet metal workers (RR 1.30), concrete and construction workers (RR 1.30), butcher and meat preparers (RR 1.33), crane or hoist operators (RR 1.38), truck or conveyor operators (RR 1.26), docker and freight handlers (RR 1.78), and launderers or dry cleaners (RR 1.79). In general, these associations were stronger among those in the same occupations in 1960 and 1970 and remained after adjusting for occupational sector. We also found increased risks for petrol station workers, textile workers, upholsterers, furnacemen, metal platers, coaters, and waiters among others.

Women

The RRs for Swedish women by two digit occupational groups are shown in table 4. In professional and technical occupational groups, significant excesses were found for health and nursing work taking as reference the same occupational sector, and non-significant excesses were found for bookkeeping and cashier's work. The RR was also increased among women employed in the production sectors, including engineering, electronics, wood, and food industries. Among service and related occupations, significant increases in risk were found for housekeeping work and caretaking and cleaning work.

Table 5 shows RRs for those specific occupations (three digit) with a minimum of 10% excess risk in women. For practical nurses, a nearly significant RR of 1.11 was found in the general cohort, whereas the comparison of these women with the rest of professionals and technicals yielded a 44% excess risk. Furthermore, the subcohort of women with this occupation in both the 1960 and 1970 censuses clearly confirmed this association. Other occupations that registered significant increases for gastric cancer were cashiers, bank employees, other engineering work, other electronic work, canning workers, plastic product workers, cooks, kitchen maids, and cleaners.

DISCUSSION

We have studied the relation between incidence of gastric cancer and occupation in a historical cohort of Swedish male and female workers gainfully employed in 1970. Risks were also estimated for the subcohort of these people who reported the same occupation in 1960 and 1970. Thus, we obtained a subgroup of long term employed workers, increasing the specificity of risk estimators.

As was expected, in the general cohort there were substantial differences in risk across occupational sectors. High risks of stomach cancer were identified among blue collar workers, whereas low risks were noted among professional and technical workers and in the administrative and managerial sector. Previous studies have shown that mortality from gastric cancer is greater in men of low social class.^24,32 Social class and educational level are strongly correlated with occupation, and it is thus difficult to disentangle the effects of occupational exposures from other lifestyle risk factors such as diet or hygiene standards.

As a way of allowing for socioeconomic status, analyses within sectors were also conducted—namely, using only workers of the same occupational sector as reference. The reduction in the risks for most of the blue collar occupations was a logical consequence of this correction. We were unable to take into account diet and other potential risk factors that could act as confounders. Some of these factors might be distributed following a geographical pattern, as incidence of gastric cancer presented a twofold variation among Swedish counties, with higher incidence in the north and lower incidence in the central part of the country. To take this variability into account, RRs were adjusted for geographical risk area.

A major problem of this type of study is that the many comparisons could produce some spurious significant results, the so called mass significance phenomenon. It is possible to calculate the number of significant results expected by chance. For this tumour, among men chance would explain three of the increased RRs that reached significance by specific occupations (three digit) and one by occupational group (two digit), whereas among women chance would account for two of the increased RRs in both three and two digit analyses.

In our data, after adjusting for age, period of diagnosis, and geographical area, a marked and consistent excess of stomach cancer was noted among workers exposed to mineral dust in the mining and masonry industries. Miners and quarrymen registered an important excess risk, something already described in Sweden.¹⁵ Interestingly, men reporting such occupation in both censuses showed a higher risk. Miners and quarrymen in this cohort have been exposed to mineral dusts of various kinds, as there are no coal mines in Sweden. The increased RRs among masonry and concrete workers reported previously,^{15,17,19,23,39} were also found in our cohort. People in these occupations have been exposed to various chemicals and dusts. Taken together, these results speak in favour of the suggested association between gastric cancer and exposure to dust.^27,40–42

After being inhaled, dust could be cleared by the lung and then swallowed, allowing potential carcinogenic compounds to take part in the precursor lesions of gastric cancer. Specifically, coal miners have been reported to have an increased mortality from gastric cancer.¹⁶ Ames found a risk excess in miners who had prolonged exposure to both coal mine dust and cigarette smoke, suggesting the interaction between these factors.⁴³ Coal miners have been exposed to very high concentrations of coal dust during their work that often are high enough to exceed the clearing capacity of the lung. More recently, a review of the risks of gastric cancer undertaken by Cocco gathered evidence around the potential carcinogenic mechanisms for coal dust and nitrogen oxides.¹⁴

Occupations associated with exposure to metal dust also showed an excess risk of gastric cancer in our study. Associations tended to be increased in the subcohort of those people who reported the same occupation in 1960 and 1970, which is consistent with previous reports.^25,41,44,45 Among men, a high risk was found for furnacemen, rolling mill workers, sheet metal workers, welders, flame cutters, metal platers, and coaters. In women, other engineering industry workers had a 50% significant excess risk in the general cohort. Exposures to several metals classified as carcinogenic to humans, hexavalent chromium, nickel, and cadmium, and to combustion products, including polycyclic aromatic hydrocarbon compounds are common in several of these occupations, and may play a part in the increased risks found.

Although male engineers did not present an increased incidence of gastric cancer when the general cohort was taken as reference, a significant excess risk was found among electrical and mechanical engineers compared with other professional and technical workers. Recently, a case-control study by Ekström et al also found non-significant increased risks for technicians and engineers when adjusting for socioeconomic status.²⁵ Parent et al¹⁹ reported an increased incidence of gastric cancer among male electrical workers who had been employed for at least 10 years as electric or electronic workers, although other studies in the United States, England, and Sweden failed to support this association.^46–48 We have also found an excess risk among women electronic workers and telephone operators. Potential exposures among these workers included electromagnetic fields and polychlorinated biphenyls (PCBs), although there is no evidence about their influence on the development of gastric cancer.⁴⁹

The relation between the rubber industry and gastric cancer is well documented.^27,31,50,51 The relevant exposures include nitrosamines and polycyclic aromatic compounds, both related to gastric cancer. In our case, a high although non-significant risk was found in the rubber industry only in men.

A high risk of gastric cancer was found for motor vehicle drivers. Excess risks have been reported for transport workers in previous studies^19,22 and recent evidence from work in Swedish men exposed to diesel engine emissions showed an increased risk of stomach cancer.³⁰ We also found an increased risk of gastric cancer among petrol station workers. Lynge et al studied the relation between cancer and exposure to gasoline vapours in the Nordic countries using the occupation registered at the census on 1970, and they found a non-significant excess risk in both sexes for stomach cancer.⁵²

Exposure to wood dust has been shown to be carcinogenic to humans, increasing the risk of nasal adenocarcinoma. Previous studies have reported a high risk of stomach cancer among these workers in Canada,²⁸ the United States,²⁹ and New Zealand.⁵³ Our data do not support this association in men, although in women wood work had a non-significant excess. In the Nordic countries, the results reported by Andersen et al were also negative.³³

A high risk of stomach cancer was found in Swedish fishermen as has been previously noted in Nordic countries.^15,33 High incidence and mortality from gastric cancer have been reported among Icelandic seamen and fishermen,^54,55 and Italian sailors and seamen.³¹ In Hong Kong, fishermen showed a high risk not only for stomach cancer but also for digestive cancers,²² suggesting the importance of dietary habits in this group. The exposures involved could be related to a high intake of fish, preserved, smoked, and cured foods, and lower consumption of fruits and vegetables.

Cancers among farmers have been widely investigated, and different studies have described an association between risk of gastric cancer and agricultural work,^{15,19–21,56,57} although neither our results nor those of some other studies supported such an association in the Nordic countries.^33,58 Differences in agricultural practices may account for these discrepancies.

Men employed in the two digit title warehouse and packaging work presented significant increases of incidence of gastric cancer, but the greater excesses appeared for its corresponding three digit category docker and freight handler. These workers are exposed to various mixed dusts and polycyclic aromatic compounds. Although not large, this association was previously described in Sweden¹⁵ and the Nordic countries as a whole.³³

Several studies have found an excess risk of stomach cancer in employees in the food industry.^19,21,59 In general, we found an increased risk of gastric cancer among them. It was significant for male butchers and meat preparers and female canning workers. The analysis for the subcohort reporting the same occupation in 1960 and 1970 confirmed the presence of an increased risk. Recently, Boffetta et al suggested that the excess of gastric cancer among butchers could be due to effect modification by lifestyle or socioeconomic factors.⁵⁹ However, in our study, when RRs were calculated with only other occupations in the same sector as reference, the excess risk, although reduced, remained increased. This results are, however, difficult to interpret, as several different factors could be involved.

In service occupations, for men significant excess risks were found among launderers and dry cleaners whereas non-significant increased risks were found among fire fighters and waiters. Among women the risk of gastric cancer was increased for cooks, launderers, and dry cleaners. Female cleaners seemed to be a high risk group in an Italian study as well as our own.³¹

In summary, the high risk among miners, construction industry, and metal processing workers would support a dust theory, as well as the increased risk among women wood workers and other engineering industry workers. The results also corroborate the strong risk among fishermen and may indicate an area that warrants further research, which could include study of the presence of toxic substances in their diet and other lifestyle factors. This study also provides support for increased risks of stomach cancer among dockers. Furthermore, there were interesting increased risks in women, for whom occupational risks have been less studied—such as practical nurses, food industry workers, cooks, and launderers and dry cleaners.