Renal cell cancer and occupational exposure to chemical agents.

A case-referent study of occupational risk indicators of renal cell adenocarcinoma was conducted. Each incidentcasein Finland in 1977-1978 was matched with two population referents. Lifelongjob histories were collectedand translated into indicators of industry, occupation, and occupational exposures. The analyses of 338setsof cases and referents revealedelevated risks for a history of employmentin white collar occupations; the printing industry; the chemical industry; the manufacturing of metal products; mail, telephone, and telegraph services; and iron and metalwarework. A decreased risk was observed for male farmers. An elevated risk and an exposure-response relationship were found for gasolineex posure. The excess risk washighest at a latencyperiod of approximately30years. The findingssupport the hypothesisthat exposureto someconstituent(s) of gasolineincreases the incidenceof renal adenocar cinomain humans. Suggestionsof elevatedrisksappearedfor exposures to inorganiclead, cadmium,and nonchlorinated solvents.

The occupational determinants of ren al ca ncer are poorly und er stood. Although several epidemiologic studies  have related specific industries and occupations with the risk of renal cell cancer, no consi stent pattern emerges from the se studies. Epidemiologic and experimental evidence has brought forth hypotheses on pr esumed work-related determinants of renal cell cancer. Those of primary concern are various hydrocarbon derivatives (24)(25)(26), petrochemical s and gasoline, jet fuel and other products of petroleum refining (16,(27)(28)(29); gasoline exhaust (18,30); cadmium (31); inorganic lead compounds (10,32); and asbestos (21,22,33).
Th e obj ective of this study was to provide further insight int o the occupational etiology of renal cell can cer.

Subjects and methods
The risk o f contracting renal cell cancer was studied in relat ion to industrial categories , occupational titles, and occup ational exposures in a case-referent study among Finnish residents ever engaged in worklife either as an employee or as an entrepreneur. All cases, 672 in number , o f primary renal adenocarcinoma (code 189,00 of the International Classification of Diseases, seventh revision) newly diagnosed in 1977 and 1978 Reprint requests to: Mr T Partanen, Department of Epidemiology and Biostatistics, Institute of Occupational Health, Topeliuksenkatu 41 a A, SF-00250 Helsinki, Finland. among Finni sh residents over the age of 20 years at diagnosis were identified th rough the Finnish Cancer Registry. Two referents, individually matched to each case on year of birth , gender, and survival status at the time of data collection, were randomly identified, within the con straints of matching, from the Population Register Centre. A questionnaire, with up to two reminders, on job history, smoking , consumption of coff ee, and obesity was sent to both the cases and the referents who were alive and to the next-of-kin of the deceased subjects. There was no reference to renal cancer in the que stionnaire. The subjects were advised to contact an interviewer by telephone if they felt that it would be more convenient for them. The number of completed forms for the cases or their next-of-kin was 461 (response rate 69 %); for the referents or their next-of-kin the corresponding number was 915 (68 0/0) (table 1).
After exclusion of the subjects who did not report any history of employment or self-employment (eg, those permanently incapacitated for work and housewives), 408 eligible cases and 819 referents remained (table 1). The number of eligible matched sets of cases and referents was 338, 192 cases having one referent and 146 cases having two referents.
The mean age, at diagnosis, of the eligible cases and their referents was 63 (range  years. The women were older than the men by 2.4 years on the average. The questionnaire requested information on the different jobs held, time periods of employment, and the identification of employers. This information enabled the reconstruction of the individual occupational historie s on an annual basis . The histories were coded, year by year , into industries and occupational cate- The individual occupational histories were then scored by an industrial hygienist (PH) into annual indicators for the years 1920 through 1968 for each of the following occupational exposures: asbestos, exposure to livestock, cadmium and cadmium compounds, lead and inorganic lead compounds, oil mist, nonchlorinated solvents, gasoline, diesel fuel and other distilled fuel oils, and polycyclic aromatic hydrocarbons (PAH). The industrial hygienist was blinded as to the case-referent status of the subjects. The scoring procedure relied on both the knowledge and experienceof the industrial hygienistand data obtained from 190interviews conducted with technical personnel, foremen, occupational health nurses, or workers who had experiencewith the processes, ventilation, raw materials, and hygienic conditions of the workplaces during the years of interest. Each annual exposure was scored on the following ordinal scale: background level of exposure, low-level exposure, high-level exposure, and exposure level not known. The definitions (table 2) of the three levels of exposure (background, low and high) were applied to the annual data of each subject. The level of exposure was classified as low or high if the levels given in table 2 were considered to exist longer than 30 % of the normal annual worktime. The concentrations defining the different levelsof exposure formed approximate guideposts rather than precise figures.
Since vapors of engine fuels are complex mixtures of a substantial number of hydrocarbons and additives, benzene was used as a convenient indicator agent for gasoline. Gasoline exposure was classifiedas high when the benzene exposure was considered to have exceeded 1 ppm and low in the approximate range of 0.1-1 ppm. Diesel fuels and related mixtures do not volatilize to a degree that is significant for human inhalation exposure except under special conditions, for example , in confined spaces or at high operating temperatures. The exposure to diesel and other distilled fuel oils was considered high when these mixtures were used as cleaning agents or were used continuously in the repair of diesel engines. Low-level exposure referred to the operation of diesel engines.
Care was taken to avoid false positive classifications of industries, occupations, and exposures, as they induce considerable bias in the odds ratio (OR) estimates when the proportion of true positive classifications is low. Uncertain situations were coded as missing values and deleted from the analyses.
For the estimation of the odds rat ios for each agent, the annual individual assignments of the different exposures were transformed into summary indicators of the no-yes type as follows: 0 =background exposure, 1= at least five years of high-or low-level exposure before 1968 or less than five years of exposure but at least one year of high-level exposure during 1920-1968, * =unknown (deleted from the analyses) or exposure categories between 0 and 1 (deleted to create an effective contrast).
Conditional logistic regression was used throughout the analysis, the case-referent status always being the dependent variable. The incidence density ratios for the different industries , job-title categories , and exposures were estimated as odds ratio s, with the respective summary indicators as regressors. The corresponding 95 % confidence intervals (95 070 CI) were calculated on the assumption of a Gaussian distribution of the estimates of the coefficients of the logit model. The analyses were done separately for the men and women, and , in some analyses, for both genders combined. Questionnaire-based indicators of smoking (never/ever) , coffee consumption (no/medium/heavy) and obesity (five-point rating scale from lean to obese) were added to most of the models to control for possible nonoccupational confounding by these variables. Missing values (16 for smoking, 17 for obesity, and 20 for coffee) were replaced by gender-specific group means for these, and only for these, variables . Coffee consumption was treated as a potential confounder, as heavy coffee drinking may be associated with renal cell cancer (29,37,38).
In addition to the qualitative analyses, exposureresponse relationships for gasoline exposure were examined. Renal cell cancer risk was analyzed in relation to (i) the mean level of exposure (estimate of the time-weighted average in parts per million of benzene in the gasoline vapor, calculated across all the years with higher than background level of exposure in the period 1920-1977), (ii) the total duration of exposure (in years) in the same period , (iii) the cumulative exposure (ppm-years = product of mean level and duration), and (iv) the latency period (1978 minus year of first recorded exposure).

Industries and occupational categories
The odds ratios associated with employment or selfemployment for at least five years in selected industries and occupations in the period 1920-1968 are summarized in tables 3 (men) and 4 (women). The selection of the industrial and occupational categories was done on the basis of a minimum of two cases in the industry or occupation. In addition, some industrial and occupational subcategories were not analyzed because we postulated that they did not involve any conceivable determinants of renal cell cancer.
Among industries, a statistically significantly depressed risk was observed for male farmers . Excess risks were observed for both the men and women employed in the printing and publishing industry, for the men employed in chemical and related industries, for the men employed in the manufacturing of metal products, machinery, and equipment, and for both the men and women in mail, telephone and telegraph services. For the mail, telephone and telegraph services, the OR pooled over the two genders was 4.9 (95% CI 1.3-18.4).
Among occupations, a significant excessrisk was associated with the aggregate category of technical art istic, administrative, managerial and clerical occupations for the men but not for the women. A statistically significant deficit occurred for the men in agricultural occupations but not for the women. Elevated OR estimates were associated with men in iron and metalware occupations and with men in graphic work. The latter excess was concentrated in the small subgroup of printers (OR 6.0) . Table 5 shows the OR estimates for the different occupational exposures for the men and women combined, unadjusted for any confounders. Gasoline exposure was significantly associated with the risk of renal cell cancer. The OR estimates for cadmium, lead, and nonchlorinated solvents were elevated, but not statistically significantly so. In the subpopulation of blue-collar occupations (white-collar and farming occupations being excluded), the OR for lead and lead compounds was 5.2 (95 % CI 0.5-6.5) as compared with 2.9 for the entire data, for non chlorinated solvents it was 4.5 (95 % CI 0.9-22.2) compared with  Among the women, the only exposure with a sufficient number of exposed subjects was the category of animal exposures. The OR, adjusted for confounding by smoking, coffee consumption, and obesity, was 1.3 (95 % CI 0.8-2.2).

Gasoline exposure
The level of gasoline exposure was classified as high (h) for 11 cases (28 % of the 39 exposed cases) and two referents (6 % of the 36 exposed referents) and as low (I) for the remaining exposed cases and referents. The persons exposed to gasoline had worked as vehicle mechanics (N h = 5), operators of a relief-printing press (N, = 3), a rotogravure pressman cleaning cylinders with gasoline (N, = 1), a weigher of containers and trucks in an oil refiner y (N, = I), a road tanker driver (N, = 1), a service station attendant (N J = 1), a firefighter (N,= 1), tax i drivers (N 1 =4), operators of other gasoline-dri ven vehicles (N 1 =22).  a Exposure = at least 5 years of a high or low level of exposure before 1968 or less than 5 years of exposure but at least 1 year 01 a high level of exposure during 1920-1968.
b Conditional logistic regression.
The simultaneous adjustment for the three exposures associated with gasoline exposure was attained through the recalculation of the OR for gasoline with linear adjustment in the logit model for exposure to diesel and other distilled fuel oil, oil mist, and animals , as well as for smoking , coffee consumption, and obesity. The OR was 1.2 (95 % CI 0.6-2.5).
Next, the exposure-response gradients for gasoline exposure were scrutinized. The sparseness of the data did not allow for the simultaneous analysis of the mean level, duration, cumulated exposure, and latency period of gasoline exposure. We therefore performed four separate conditional logistic regressions entering one exposure dimension at a time. The distribution of the cases and referents along the four dimensions was unevenly spaced . To increase the stability of the esti- mates, we classified each exposure dimension into four categories containing a reasonably balanced number of subjects. The four categories of each dimension are defined in table 8. Each dimension was treated as a factorized predictor, the unexposed subjects forming the reference category. This procedure allowed for the simultaneous estimation of three OR estimates contrasting the three exposure categories against the unexposed reference category. Obesity, smoking, and coffee consumption were included in each model. Table 8 shows that the risk of renal cell cancer increased with increasing level of gasoline exposure, as approximated by the benzene-equivalent concentration. The highest level (about 1.0-2.0 ppm of benzene) was associated with a significantly elevated OR of 7.4. No clear trend in the risk was noted for the duration of exposure. The cumulative exposure, reflecting both the level and the duration of exposure, displayed a less steep upward gradient than did its constituent, the level of exposure. The OR for the highest category (benzene equivalent of about 14-102 ppmyears) was 4.3. The latency category associated with the highest OR was that of 27-33 years.

Discussion
This report addresses the risk of renal cell cancer over industries, occupations, and occupational exposures. No strong ad hoc hypotheses of the chemical exposures 236 as determinants of kidney cancer were set, with the possible exception of exposures to gasoline constituents and lead compounds.
The finding of an increased risk for men in the mixed category of different white-collar occupations is in agreement with results of a number of studies (1)(2)(3)(4). The occupational exposures considered in this study were uncommon in the white-collar occupations. There was no clustering of cases into particular subcategories of industry or occupation among the men in whitecollar occupations. Obesity, coffee consumption, or smoking did not explain the elevated risk .
In concordance with Danish (1) and Swedish (2) experience, a significant deficit in renal cancer risk was observed for male farmers.
The excess risk associated with employment in printing and publishing may be explained by gasoline exposure, which was fairly common in this industry, particularly among the men. In addition , exposure to lead and inorganic lead compounds was more frequent for the subjects in this industry than for the rest of the subjects . Both lead and, in particular, gasoline were associated with the risk of renal cell cancer, a finding in agreement with those of animal studies (27,32). A related epidemiologic finding is that of Paganini-HiII et al (20), who reported a significantly increased mortality from kidney cancer among newspaper web pressmen.
An increased risk was associated with the manufacturing of metal products, machinery, and equipment and iron and metalware work among the men . The exposures characterizing these jobs were cadmium, lead, oil mist, solvents, and PAH.
Men in the chemical, petroleum, coal , rubber, and plastics industries had an elevated risk but did not display anything unusual in the frequency of the exposures considered in this study. Closer scrutiny of this heterogeneous group revealed a cluster of three cases out of eight (but none among the referents) with a history of employment in the petroleum industry.
An excess risk was significantly associated with employment in mail, telephone, and telegraph services. No exposure under study was prominent in this industry. The represented occupations comprised postal van drivers, telephone and line installers, and various office jobs.
An elevated OR was observed for cadmium exposure among the men, but it was based on three cases onl y. The scanty prior evidence is restricted to a single epidemiologic study by Kolonel (31), who reported a significantly elevated OR for occupational exposure to cadmium in a case-referent study.
Lead acetate, lead subacetate, and lead phosphate have induced renal tumors in rats and mice by several routes of administration (32). Selevan and her collaborators (10) have reported elevated kidney cancer mortality for workers in a lead smelter. Two cases of renal cancer associated with heavy exposure to lead have also been reported (39,40). Malcolm & Barnett (41) observed an excess of deaths due to renal diseases among workers exposed to lead in factories producing lead acid batteries. Renal neoplasms were not mentioned by the authors. We found an elevated but statistically nonsignificant OR for exposure to lead and inorganic lead compounds. Of the high-risk industries and occupations, significantly elevated proportions of prior exposure to lead compounds was observed in printing and publishing and iron and metalware work. These findings are somewhat imprecise and do not therefore add much to the existing evidence for the hypothesis of an association between (inorganic) lead and renal cancer. The fact that the gasoline used in Finland, especiallyin the past, contains tetraethyllead may be relevant in that gasoline exposure was associated with renal cancer risk . Tetraethyl lead is capable of dermal entrance to the organism, where a part of it is transformed into inorganic lead compounds (42). The amount of lead intake is probably low in the jobs with typical dermal exposure to gasoline, however.
Asbestos exposure has been related to an increased risk of kidney cancer in three epidemiologic studies (21,22,33). Asbestos fibers have been detected in the kidneys of exposed subjects (43,44). Our results fail to substantiate these findings. The false negative rate of exposure assessment was probably highest for asbestos in our study. In addition, it is possible that the latency periods were insufficient, as the peak of asbestos use in Finland occurred around 1970. It would therefore seem that our OR estimate for asbestos was negatively biased. To the contrary, however, occupational titles with a high probability of asbestos exposure (dockyard workers, asbestos insulators, and workers in asbestos quarries, mills and asbestos product manufacturing facilities) were completely absent from our data. Our finding on asbestos is inconclusive.
We observed an elevated OR for nonchlorinated solvents for men . These solvents were constituents of paint and glue formulations (eg, toluene, xylenes, ketones, esters, alcohols, benzene, mineral spirit, and turpentine). The number of persons exposed to chlorinated solvents was too small for statistical analysis. It might be noted that Duh & Asal (12), Katz & Jowett (13), and Asal et al (29), who found an elevated kidney cancer risk among laundry and dry-cleaning workers, did not discriminate between petroleum-based and chlorinated solvents as possible risk factors. There was only one case in our data with a history of dry-cleaning work (as an operator).
Gasoline exposure was consistently related to kidney cancer risk in our data. It persisted after control for obesity, smoking, coffee consumption, concomitant diesel fuel exposure, social status, and farming occupations. An exposure-response relation was observed for the level of exposure to gasoline, and a weaker such relation was seen for the cumulative exposure (product of mean level and duration). Duration of exposure alone was not linearly associated with the risk of renal cancer, a finding which might reflect 2 the correlatedness between latency and duration, the latency peak appearing at the intermediate category of around 30 years. The composition of the imported gasoline used in Finland in the late 1940s and early 1950s is not well known . Tetraethyllead was used as an anti-knock additive.
Regarding the assessment of some individual exposures to gasoline, the printing press operators (three cases and a referent) used gasoline as a cleaning agent in typographic printing shops. This use was confirmed by nurses and safety personnel of the printing shops concerned. The gasoline exposure (classified as low level) of a firefighter was based on the reporting of the case himself. The exposure of the operators of gasoline-driven vehicles, taxis included, was classified as low level, as their gasoline exposure was considered higher than in the population at large.
Our results suggest that some hydrocarbon constituent(s) or additive(s) of gasoline are conducive to renal cell cancer in humans. The level of exposure is probably important, and the postulated average latency period was about 30 years for the exposure conditions encountered in this study. Epidemiologic studies conducted elsewherehave, however, failed to produce convincing evidence of a relationship between gasoline exposure and renal cell cancer, with the possible exception of aviation gasoline reported by Siemiatycki and his co-workers (28). A suggestion of an exposureresponse relationship between the duration of employment as a gasoline station attendant and the incidence of renal cell cancer has been reported by McLaughlin and his co-workers (23). The implicated dose-response relationship between unleaded gasoline and kidney cancer for the male rat (27) suggests some gasoline hydrocarbon(s) as responsible for the conjectured causal association. However, this finding may also be due to a male rat-specific protein (alpha-2-microglobulin) in the kidney, accumulated in the male rat as a result of exposure to a nephrotoxic hydrocarbon (2,2,4-trimethylpentane), a typical component of gasoline (45)(46)(47). The role of lead compounds cannot be ruled out when human exposure is under consideration.
Some remarks on the validity of the study are warranted. The misclassification rates of exposures were probably nondifferential between the cases and the referents because the coder was blinded as to the casereferent status of the subjects. Since the subjects were matched for gender, age, and vital status, the probability of differential misclassification of exposures due to the matching factors (eg, old age) should be low. Every effort was undertaken to avoid, in the scoring of the exposures, false positive misclassifications of exposures, which would have been particularly prone to bias the OR estimates towards unity . High rates of false negative misclassification would however have the same effect, which may be reflected in our result for asbestos exposure. Some nondifferential misclassification may however have been possible by random 237 fluctuation when the number of exposed persons was small.
Nonresponse would bias the effect estimates if it were differentially correlated with industries, occupations, and exposures between the cases and referents. This possibility could not be directly checked. We cannot however envisage any reason why the referents exposed, for example, to gasoline would tend to refrain from replying, as compared with the cases. Insofar as nonresponse and the accuracy of work histories are dependent on age, gender, and the data source (self or next-of-kin), the matching that was performed should preclude information bias from these sources. The exclusions of subjects due to incomplete case-referent sets in the conditional analyses were in all likelihood randomly distributed between the cases and referents in occupational categories and exposures. We therefore believe there was no serious bias introduced by these exclusions. The exclusion of all subjects with no reported occupational history was a deliberate restriction of the source population, as those without such a history were noninformative in the study base, which was designed to address occupational exposures.
In conclusion, the most important finding of this study is the qualitative and quantitative evidence contributing credence to the hypothesis that exposure to some constituent(s), hydrocarbons, or additives of gasoline increases the risk of renal adenocarcinoma in humans. Our data do not allow for further chemical specification of this hypothesis.