Occupation and the risk of lung cancer in Uruguay

Objectives The purpose of this study was to provide more information regarding the risk of lung cancer associated with asbestos, pesticides, and other exposures in the Uruguayan work force. Methods This multisite case-referent study was part of a large project designed for evaluating the role of occupational exposures in cancer risk in Uruguay. According to the design employed, cases were a subset of the data base conesponding to a particular site (in this instance I~ing cancer), and they were compared with all other sites conibined (referents). Rt?Slllfs Significant increases in risk associated with worlters in the construction industry were inainly observed for squamous-cell carcinoma. Asbestos, silica dust, and DDT (dichlo~~odiphenyltrichlosoetllane) exposure were also associated with increases in the ~isk of lung cancer. Pipefitters, bakers, and textile workers were also at increased risk of developing lung cancer. C O ~ C ~ ~ S ~ O ~ S Workers einployed in the constructio~l industry, as well as those exposed to DDT may have an excess risk of lung cancer. These findings are particularly important in showing that developing countries like Uruguay display risk patterns of similar magnitude as those observed in developed conimunities.

Lung cancer is the most frequent malignant disease among the male population of Uruguay with an incidence rate of 74.0 per 100 000 (1). Mortality from lung cancer in Uruguay is among the highest of the American countries, following closely the United States and Canada, and it is increasing with time (2)(3). Previous studies in Uruguay have shown the donlinant role of tobacco s~n o k i~l g in the etiology of the disease, with an attributable risk of 85% (4). In a previous case-referent study (5) it was estimated that about 22% of all lung cancer could be attributed to occupational exposures, and a higher risk of lung cancer was observed among construction worlters, welders, and electricians. This finding suggests that exposure to asbestos may be widespread among the Uruguayan work force. In addition, agricultural workers displayed a twofold excess in risk for small-cell carcinoma of the lung. Since agricultural worlters represent about 20% of the employed population in Uruguay, this country constitutes a convenient place for studying the possible association between lung cancer risk and pesticide exposure. We conducted a case-referent study designed specifically to describe the possible role of exposure to asbestos, pesticides, and other occupational exposures in the risk of lung cancer in Uruguay.

Subjects and methods
This study was part of a large multisite case-referent study aiming to estimate the importance of different occupational exposures for the occurrence of cancer in Uruguay. All incident cases occurring in men between 30 and 75 years of age and admitted to the five major hospitals in Montevideo, Uruguay, in the period January 1993 to December 1994 were interviewed shortly after admittance by a team of trained social workers. All sites and histologies were considered eligible for the study. The response rate was high (97.4%), and the patients were extremely cooperative. The reasons of nonresponse were refusals (2.0%) and poor physical condition (1.6%). Accordi~lg to data from the National Cancer Register, iilterviewed cases represeilted 40.0% of all the incident cases registered in the same period in Montevideo (43% of the cases and 36% of the referents).
The distribution of cases by cell type is shown ill table 1. Squamous-cell carcinoma was the most frequeilt (50.7%), followed by adenocarciiloina (21.1%), and small-cell lung cancer (12.2%). In table 2, the referents are classified by cancer site (International Classification of Diseases, ninth revision). Colorectal cancer was the most common site (30.6%), followed by prostate (23.7%) and malignant lymphotna (8.1 %).
The questionnaire covered sections on sociodemographic variables, tobacco smoking, alcohol drinking, a complete occupational history, specific questio~ls concerning the use of several substances, an asbestos section including relevant job titles and probing questions for each job title, and a section on pesticide use for agricultural workers and garden owners. The part of the questionnaire evaluating exposure to asbestos was an abridged version of a series of suppletnentary questiontlaires used in a German study on risk factors for lung cancer (6).
For each occupation held for at least one yeas, age of employme~lt, year of hire, and duration of the exposure was recorded. The subjects employed in each occupation were coinpared with subjects never employed in the corresponding occupations.
Furthermore, estimates of exposure to specific substances were carried out for each patient, according to the principles previously employed by Jocltel et a1 (6). According to the strategy proposed, each cancer site is examined as a case series, using the remaining cancer sites as referents.
In the present study lung cancer patients constituted the case series. The re~nailling sites, with the exceptioil of cancer of the oral cavity, pharynx, esophagus, stoinach, larynx, and bladder, constituted the reference series. The reason for excludiiig these sites in the reference group was the anatomic proximity or the likelihood of shared occupational etiologies. A total of 270 cases and 383 referents were included in the study. The age distribution was slightly wider for the referents. A larger proportion of lung cancer patients lived in the capital city of Montevideo, and a smaller percentage lived in the 1101thern region of the country. The cases were somehow less educated and with lower incomes than the referents, but the differences were nonsignificant.
Relative risks, approxiinated by the odds ratios (OR), were estimated for each job title and substance, after control for major confounders, through stratified analysis and unconditional logistic regression (7). The decisions as to which covariates would be included in the final models were based on the biological plausibility, the level of significance when entered into the model (at 0.10) and confounding (whether the covariate acted as a confounder of the primary association of interest).
(Confounding was co~lsidered present if the regression coefficieiit of the priinary independent variable changed more than 10% after the addition of the potentially confounding variable to the model.) Ainong the potential confounders, age, residence, education, income, tobacco smoking (in pack-years), and alcohol consumptio~l fulfilled the critcria and were included in all the models. More complex models included intcraction terms between job titles or substailces with tobacco sinokiilg and sociodernographic variables. All the calculatioiis were perforined using EGRET and GLIM progralns (8)(9).
The odds ratios for selected job titles, after adjustment for major confounders, is shown in table 3. The only job titles associated with sigilificantly elevated odds ratios were briclilayers (OR 1.6) and finishers (OR 1.6). Increased risks were also observed for butchers (OR 2.0), shoemakers (OR 4.8), pipefitters (OR 1.9), textile workers (OR 2.3), and plumbers (OR 2.2). There was a significant effect for textile worliers with duration of employment; longer duration of textile work being associated with a fourfold increase in risk. When job titles were classified by cell type, importailt differences resulted (table 4). Significantly elevated odds ratios for construction workers (bricltlayers, fillishers) were associated with  Table 3. Relative risks of lung cancer for job titles with at least five cases exposed to asbestos. (OR = odds ratio, 95% CI = 95% confidence interval)

Discussion
The results of our study indicate several associations between job titles and substances and the risk of lung cancer. In fact several hazardous industries are present in developing countries, where the risk is increased by the lack of protective measures (10). This situation is not unknown in Uruguay. In addition, Uruguay is a major cattle and wool produccr and occupations like farming are of outstanding importance and allow an assessment of the effects of pesticide exposure in a large sector of the population of Uruguay. Several findings should be emphasized. Asbestos is a well defined lung carcinogen (1 1-14), and our results show that 26% of the study population was exposed to asbestos. Furthern~ore, the risk appears to be largely confined to squainous-cell cancer. This finding appears to be in accordance with the findings of Baker et a1 (13), who reported a higher frequency of this cell type among workers exposed to crocidolite.
The subject has been reviewed recently by Vainio & Boffetta (1 1). Unfortunately, the type of asbestos fiber to which this population was exposed is rather difficult to assess, given the design employed. Other studies have reported an increased frequency of small-cell lung cancer in relation to rnixed and cornplex exposures to asbestos (13)(14).
Silica dust is a probable human lung carcinogen (15). Our results suggest an elevated risk for squamous-cell carcinoma of the lung. Siemiatycki et a1 (16) were unable to find an association between silica exposure and lung adenocarcinoma, squamous-cell and small-cell lung cancer being the only types associated with this substance. Wood dust (years) 1-20 years of exposure 2 21 years of exposure Ever exposed Mineral wool dust (years) 1-20 years of exposure 21 years of exposure Ever exposed Fur dust (years) 1-20 years of exposure 2 21 years of exposure Ever exposed Gasoline (years) 1-20 years of exposure 2 21 years of exposure Ever exposed Diesel exhaust (years) 1-20 years of exposure r 21 years of exposure Ever exposed Gasoline exhausts (years) 1-20 years of exposure r 21 years of exposure Ever exposed Excavation dust (years) 1-20 years of exposure r 21 years of exposure Ever exposed Portland cement (years) 1-20 years of exposure 2 21 years of exposure Ever exposed Brick dust (years) 1-20 years of exposure r 21 years of exposure Ever exposed Solvents (years) 1-20 years of exposure 2 21 years of exposure Ever exposed Asbestos (years) 1-20 years of exposure r 21 years of exposure Ever exposed Silica dust (years) 1-20 years of exposure r 21 years of exposure Ever exposed Arsenic (years) 1-20 years of exposure 2 21 years of exposure Ever exposed DDT (dichlorodiphenyltrichloroethene (years) 1-20 years of exposure r 21 years of exposure Ever exposed Dieldrin (years) 1-20 years of exposure r 21 years of exposure Ever exposed Other and unknown pesticides 1-20 years of exposure r 21 years of exposure    Other important associations observed in our study were those with Portland cement and brick dust, and for both substances the higher risks corresponded again to squamous-cell lung carcinoma. Nevertheless, the lack of statistical power assessiilg associations with small-cell carcinoma and adenocarcinoma should be taken into account. Both exposures are of great interest, given the high proportion of constructioil workers in our base population (29.7% of the reference series). These findings congrue with previous results (17).
Pesticides formed one of the main exposures of interest in this study. Our results support an increased risk of lung cancer in general, and of lung adenocarcinoma and small-cell lung cancer in particular, for DDT exposure, and the dose-response pattern was significant, replicating previous findings (18)(19)(20)(21)(22)(23). Since the pesticides used in this work force were partly unknown, it is difficult to establish which chemical could be responsible for the increased risk associated with the category of other pesticides. The working group of the International Agency for Research on Cancer concluded that DDT is possibly carcinogenic to humans (group 2B) (23). Nevertheless, the association between lung cancer and DDT exposure should be evaluated cautiously.
We have evaluated the possibility of bias in the results of this study. It is unlikely that selection bias may have considerably affected these findings, since all patients admitted to the major hospitals of Montevideo were included in the study.
We also compared the proportion of lung cancers in this base population with the proportion observed in data of the National Cancer Register. The figures were siinilar and suggest a common catchment area (28.3% in this study, 22.9% in the National Cancer Register).
Recall and interviewer biases appear to have been minimized by the design enlployed in this study. All the patients interviewed were cancer patients, and recall of the exposures under study should have been similm for each cancer site (24)(25).
As aptly stated by Siemiatycki et a1 (15), the main inconvenience of using a cancer reference group is the possibility that the cancers in the reference pool share occupational exposures with the case series and thus dilute a real effect. This possibility was minimized by including all cancer sites with the exception of sites previously considered possibly related with one or more important exposures relevant to the etiology of lung cancer. Finally, our design precluded a better characterization of the exposure, the result being a critical evaluation of the estimates presented in this study.
For a disease like lung cancer, residual confounding from tobacco smoking is always a matter of concern. We tried to minimize this problem by including a term for pack-years in all the analyses. We tried to measure the degree of confounding by tobacco smoking using the confounding risk ratio (CRR), as suggested by Spinelli et al (26). This ratio ranged from 0.70 for mineral wool fiber to 1.30 for concrete dust. In conclusion, this study has provided a rather large number of associations between occupational exposures and lung cancer, and it suggests that smaller previous studies conducted in Uruguay were valid (5).