Mortality of filling station attendants.

OBJECTIVES - Gasoline contains established human carcinogens, such as benzene. The health impact of exposure to this fuel, however, has not been fully elucidated. We report on the mortality of a co hort of 2665 filling station managers from the Latium region (Italy). M ETHODS - This is the first workplace-based cohort of gas station attendants. However, only self employed individuals were available for study (about 50% of the whole work force). The follow-up period extended from 1981 through 1992. The mortality experience of the cohort was compared with that of the regional population. RESULTS - The overall analysis for standardized mortality ratios (SMR) showed a significantly de creased mortality from all causes, mainly due to a deficit of cardiovascular diseases and malignant neoplasms. Nonsignificantly increased risks for esophageal cancer [SMR 241, 90% confidence inter val (90% CI) 82-551], brain cancer (SMR 195,90% CI 77-401) and non-Hodgkin's lymphoma (SMR 173,90% CI 47-448) were found for the men; mortality due to lung cancer and leukemia was lower than expected, and no kidney cancer death was recorded. Among the attendants of small sta tions (characterized by a small numberof employees and high sales of gasoline per full-time employee), the SMR values for esophageal cancer (for men SMR 351, 90% CI 120-803) and brain cancer (for men and women SMR 266, 90% CI 105-559) showed increased values. CONCLUSIONS - Filling station attendants are exposed to gasoline vapors and seem at risk of cancer of various sites. Due to the power limitations of this study, however, a precise estimate of the risk for many causes of death was not achievable. Further cohort studies of greater size are warranted.

Gasoline is a complex and highly variable mi xture of aliphatic and aromatic hydrocarbons, mainly of low molecul ar weight (I ). Exposures to gasoline fuel and to exhau sts from engines operating on gasoline are consid ered by the International Agency for Research on Cancer as possibly carcinogenic to humans (2)(3).
So far , how ever, the health impact of exposure to petroleum fuel s has not been fully elucidated (4) . In particular, the hazard posed by exposure to benzene via gasoline vapors and exhausts is of major concern, due to the great number of workers employed in petroleum di stribution trades and to the relevant contribution of su ch sources to the pollution burden in urban en vironments (5,6).
While in the United States and northern Europe self-service stations dominate, Italy and other southern European countries still employ large numbers of workers in ga soline retail trad es. In thi s paper we National Health Institute, Rome, Italy. Epidemiological Unit, Latium Regional Health Authority, Rome, Italy. 3 Department of Neurology, "La Sapienza" University, Rome

Subjects and methods
Coh ort design A nationwide survey of service stations was carried out in Italy in 1980. Information on several characteristics of the station s and their managers was collected during the survey, such as the name, date of birth and length of employment of the manager, the number of attendants, the quantity of fuel sold in the pre vious four years, and the number of fuel deliveries to the station per year. Detailed data for the Latium reg ion (central Italy) were obtained, and managers who were clearly identifiable as individuals (as opposed to commercial houses) and who were still alive on 1 January 1981 were con sidered eligible for the study [2665 stations and managers out of 3272 orig inal records (81.4%)].
The cohort consisted of 2665 service station managers (2308 men and 357 women) . At entry into the study their age averaged 46 years, their mean length of employment wa s 14 years, and most of them (91%) were full-time employees (table I ). At the end of the study, 2236 subj ects were still alive (84 %), 270 had died (10%), and 159 had not been traced (6%) . No information on the cause of death was available for 12 decea sed persons (4%). Some characteristics of the 266 5 service stations in 1980 are described in table 2 by statio n size. Large stations, as expected, had higher sales of gasoline, underground tanks of greater capaci ty, and a greater number of emplo yees than the small stations. The latter , however, were characterized by significantly higher sales of gasoline per full-time employee.
Each subject was considered from I January 1981 through 31 July 1991 if resident in Rome at the time of entry into the study, or through 30 June 1992 if resident in any other municipality within the Latium region. Vital status was ascertained through the registry of the last municipalit y of residenc e. The code for the underlying cause of death in the Internat ional Classification of Diseases (ICD), IX revision, was 332 retrieved by means of various method s, as described in detail elsewhere (7) : first, throu gh record linkage with the National Mort ality File for deaths occurring before 1986; second, through record linkage with the Regional Death Index for deaths occurring during the period 1986-1992; third , from the registry of the municipality where the death occurred for those cases for which linkage procedures were unproductive. In the last case [75 out of 270 deceased indi vidual s (28%)], the underlying cause of death was coded by an expert nosologist.
For the calculation of person-years at risk of dying, each subj ect was considered from I January 1981 until the end of the study period or the date of death. Those lost to follow-up were consid ered alive at the end of the study period . Standardized mortality ratios (SMR) and their 90% confidence intervals (90% CI) were used to compare the mortality exper ience of the cohort with that of the regional population. The expected number of death s was calculated with the use of gender-, age-and period-specific regional mortality rates. The statistical package OCMAP (8) was used for the analysis.

Exposure assessm ent
Informat ion on duration of emplo yment at entry, along with a series of indicator s of work load, were available for all cohort member s from the 1980 survey. No vapor recovery system has been enforced in Italy, and the relative quantity of fuel sold has remained relatively stable. Unleaded gasoline, in fact, still accounts for less than 5% of the total amount of marketed fuel. Due to the characteri stics of Italian service stations, the yearly average quantit y of gasoline sold per full-time employee can be considered an indicator of the averag e intensity of exposure experienced by the manager at refuelling.
Exposure to aromat ic hydrocarbons among currently employed filling station workers was estimated by means of an environmental survey taken in 1992 (9). Measured concentrations of benzen e, toluene , and xylenes [8-h time-weighted average s (TWA)] averaged 0.55,0.71 , and 0.32 mg · rrr", respectively. The number of vehicles filled, the daily sales of super premium gasoline and motor-bike fuel, and the winter season were all significant predictors of the logarithmic concentrati on of benzene (8-h TWA) in the simple regre ssion analyses . The size of the station acted as an exposure modifier. While no single variable was able to predict the benzene level in large stations, for small stations an increase of 0.0579 and 0.04 18 in the log benzene concentration per unit of increase in super premium gasoline (100 I) and in motor-bike fuel (l0 I), respectively, was estimated. The overall variance explained by the multivariate Scand J Work Environ Health 1994, vol 20, no 5 model was, however, only 12.3%. Therefore, a clearly defined categorization of the subjects into groups with internally homogen eous and significantly different exposure levels was not feasible. As regards the exposure assessment of the cohort, it seems that workers in small-size stations with higher sales of super premium gaso line may experience higher levels of exposure. We decided, consequently, to examine in detail this group of workers.

Results
All-cause mortality was significantly decrea sed, mainly due to a deficit in cardiovascular diseases and malignant neoplasms. The risk of blood disease (among both the men and the women) and injuri es (only among the men) was increased, but the excess was not statistically significant (table 3).
Although cancer mortality was lower than expected, nonsignificantly increased risks were found for non-Hodgkin' s lymphoma and for esophageal and nervous system cancers among the men (table 4). One case of soft-tissue sarcoma and one case of melanoma were observed versus 0.2 and 0.7 expected, respectively . The risk of colon and bladder cancers was slightly elevated, while mortality due to lung cancer and leukemia was lower than expected. No kidney cancer death was recorded , whereas two cases were expected. The number of deaths recorded among the women was very small (20 deaths, 5 cancers). Excess risks were found for pancreatic, uterine, and brain cancer, but each excess was based on one observed case (table 4).
Among the attendants of small stations, the SMR values for esophageal and brain cancer showed increased values when compared with the overall findings , reaching statistical significance among the men and in the whole subcohort. Moreover, the ex- Table 3. Mortality of the service station attendants from the Lat ium region (1981-1992) by broad groups of causes of death.

Men
Women Scand J Work Environ Health 1994, vo120, no 5 Table 4. Cancer mortality of the service station attendants from the Latium region (1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)  cesses of blood diseases, soft tissue sarcoma, and Discussion melanoma were concentrated into this subcohort This study represents the first workplace-based co- (table 5). For these workers, furthermore, an inhort of filling station attendants. In fact, epidemiocreased risk of laryngeal cancer was recorded. The logic investigations of this occupational group are excess risk of non-Hodgkin's lymphoma was, howparticularly difficult to carry out because managers, ever, almost equally distributed among the two subalthough directly involved in refueling, are self-emcohorts (tables 5 and 6). Attendants of large service ployed workers, attendants are scattered among a stations, on the other hand, showed increased risks large number of very small stations, and neither of of bladder cancer, leukemia, and nervous system these groups are generally subject to any form of cendiseases (table 6).
tralized personnel record-keeping requirements. Table 6. Mortality of the atte ndants of large service statio ns from the Latium region (1981-1992) by sele cted causes of death .
(0 = observed death s, E = exp ect ed deat hs , SMR = sta ndard ized mortality rat io , 90% CI = 90% co nfidence inte rval). The followi ng three issues are worth discussing: (i) the completeness of the cohort in relation to available sources which cata logue workers employed in this trade , (ii) a comparison of our findings with results from available relevant studies, (iii) and, finally, the major shortcomings of this study, In the 198 I Italian census , the economic sector "retail sales of fuels and lubricants" listed 32 034 local units and 60 869 workers . More than 50% of the work force consisted of self-e mployed workers (N =32 606), 20% were employed family members (N = 12439), 4% were salaried clerks (N =2470), and 22% were workmen (N = 13 354) (10). Within the Latium region, according to the census (I I), there were 3077 service stations , employing 5784 workers. Of these, 30 18 were self-emp loyed workers (54%), 1040 were employed fami ly members (18%), 172 were salaried clerks (3%), and 1464 were workmen (25%). The cohort data source recorded a higher number of service stations (N = 3272) one year earlier, a finding compa tible with normal marke t fluctuations.
We studied, therefore, about half of the whole work force employed in gasoline service stations in the regio n, namely, the self-employed managers, excluding family and salaried workers. The mortality experience of these populations might be different from that of the one under study. However, as most of the cohort members declared full-time employment [2 169 out of 2379 with available infor mation (91.2%)] and the majority of their stations employed one or two worker s [1827 out of 258 1 with available information (70.8%)], we deemed that our "filling station attendan ts" were directly involved in refue ling.
We found excess risks, although not statistically significant, for blood diseases, esophageal cancer, nervous system cancer, and non-Hodgkin' s lymphoma in the whole coho rt. Moreover, an increase in the SMR values for these causes of death was found for the managers of small stations, along with an excess risk for laryngea l cancer. Managers of large stations, on the other hand, showed increased risks for bladder cance r, leukemia, and nervous system diseases. The exposure profile of small station workers can be the most easily characterized in terms of gasoline vapors and exhausts, while attendants in large stations were involved in other tasks besides refueling, such as mechanical, electrical, and tire repairing jobs. This differe nce cou ld exp lain, at least partia lly, the different patterns of mortality obse rved.
Previous findings relating to risk exces ses among filling station attenda nts stem from surveillance systems of occupational mortality or cancer incide nce. Auto service and gaso line station workers in California (1959California ( -1961 showed significantly elevated proportio nal mortali ty ratios (PMR) for oral cavity, lung, and testicular cancers (12) . Among the service statio n owners and attendants in Washingto n State ( 1950-1979) elevated PMR value s were found for cancers of the oral cavity, kidney, bladder, and testis and for aplastic anemia ( 13). In Massachusetts (1971Massachusetts ( -1973 a small increase in the PMR for leukemia was found for auto mechanics and gasoline station attenda nts (14). Service station workers in New Hampshire (1975-1 985) showed increased mortality for leukemia, suicide, emphysema, and mental and psychoneurotic conditions ( 15). Owners of filling stations in Denmark, as defined on the basis of the 1970 Danish census, showed an increased incidence of kidney cancer over the period 1970-1980 (16) . They were also found to be at increased risk of dying from respiratory cancers over the period 1970-1987 (17) . A significantly increased standardized incidence ratio (SIR) for acute myeloid leukemia was found amon g Swedish gasoline station attendants, as of the 1970 census, followed-up over the period 1971-1984 (18). In the Montreal multi site, multiple-exposure case-referent study, exposure to automotive gasoline was associated with a significantly increased risk of stomach cancer, with a dose-response effect (19), while exposure to gasoline exhausts proved to be associated with excess risk of rectal cancer, also with a dose-response effect (20).
Our findings essentially confirm previou sly reported excess risks, with the remarkable exclusion of kidney cancer, for which no death was recorded. The deficit in mortality from all causes, cardiovascular diseases, and respiratory diseases could be ascribed to the relatively low duration of the follow-up, and to the combination of factors generally referred to as the "healthy worker effect." In spite of an increase in mortality from lung cancer, as reported in association with exposure to exhausts from engines operating on diesel fuel (3), we observed an SMR below the null value. The observed excesses of blood and nervous system diseases deserve attention, in view of the exposure to the hematological and neurological toxicants present in gasoline vapors.
A major drawback is that the power of this study, for most of the causes of interest, was low and the duration of the follow-up might have been insufficient for cancers that have a long induction period. Furthermore, becau se of the lack of information on employment termination dates , we did not examine the observed excess risks by length of employment, a usual proxy for duration of exposure.
Lack of information about smoking habits and alcohol consumption is undoubtedly a weakness of this study . Nevertheless, mortality due to tobacco-and alcohol-related causes of death (lung cancer, respiratory diseases, and dige stive diseases, particularly cirrhosis of the liver) was not increased in the co-336 hort. Alcohol consumption is of major concern in relation to the observed increased risk of esophageal tumor s. Howe ver, in the sample of current filling station attendants (9), only 27.3 % of the subjects drank one glass of wine or more per day. In the Italian population in 1983, among men aged 30->60 year s, there were 52.5 % drinkers of a quarter liter or more of wine (51). Therefore, confounding from alcohol con sumption, in respect of the association between exposure to gasoline vapors and esoph ageal cancer, seems unlikely.
In conclusion, filling station attendants are exposed to gasoline vapors containing aromatic hydrocarb ons , including benzene. This group of workers seems at risk of cancer at various sites. The observed increa sed risks are consistent with the hazards to which they are exposed. Due to the power limitations of this study , a precise estimate of the risk for many causes of death is not achievable . Further cohort studies of greater size are warranted.