Document oN2Z18exR0ez4JL8Q7oRManR

Cancer Causes Control (2008) 19:1323 DOI 10.1007/s10552-007-9065-x ORIGINAL PAPER Increased risk of acute myelogenous leukemia and multiple myeloma in a historical cohort of upstream petroleum workers exposed to crude oil Jorunn Kirkeleit Trond Riise Magne Bratveit Bente E. Moen Received: 16 May 2007 / Accepted: 29 August 2007 / Published online: 29 September 2007 Springer Science+Business Media B.V. 2007 Abstract Benzene exposure has been shown to be related to acute myelogenous leukemia, while the association with multiple myeloma and non-Hodgkin lymphoma has been a much-debated issue. We performed a historical cohort study to investigate whether workers employed in Norway's upstream petroleum industry exposed to crude oil and other products containing benzene have an increased risk of developing various subtypes of hematologic neoplasms. Using the Norwegian Registry of Employers and Employees we included all 27,919 offshore workers registered from 1981 to 2003 and 366,114 referents from the general working population matched by gender, age, and community of residence. The cohort was linked to the Cancer Registry of Norway. Workers in the job category ``upstream operator offshore'', having the most extensive contact with crude oil, had an excess risk of hematologic neoplasms (blood and bone marrow) (rate ratio (RR) 1.90, 95% confidence interval (95% CI): 1.193.02). This was ascribed to an increased risk of acute myelogenous leukemia (RR 2.89, 95% CI: 1.256.67) and multiple myeloma (RR 2.49, 95% CI: 1.215.13). There were no statistical differences between the groups in respect to non-Hodgkin lymphoma. The results suggest that benzene exposure, which most probably caused the increased risk of acute myelogenous leukemia, also resulted in an increased risk of multiple myeloma. Keywords Benzene Cohort studies Hematologic neoplasms Leukemia Multiple myeloma J. Kirkeleit (&) T. Riise M. Bratveit B. E. Moen Section for Occupational Medicine, Department of Public Health and Primary Health Care, University of Bergen, PO Box 7804, Bergen 5020, Norway e-mail: Jorunn.Kirkeleit@isf.uib.no Introduction Benzene is known to cause leukemia [1, 2]. The association is strongest for acute myelogenous leukemia, but unclear for the other subtypes of leukemia [2]. Further, the association of benzene exposure with multiple myeloma [3, 4] and non-Hodgkin lymphoma [58] has been much debated. Benzene is a natural component of crude oil and natural gas, and is a potential hazard in the petroleum industry. Studies of petroleum workers have shown an association between benzene exposure and leukemia [912]. Some studies of petroleum workers have found an elevated risk of multiple myeloma [1315], while others have not [9, 12, 16]. Further, although a relation between non-Hodgkin lymphoma and benzene exposure has been suggested [7, 8], most studies in the petroleum industry have not found any increased risk [5, 17]. The majority of these studies comprised of workers employed in oil refinery operations and the distribution and consumption of refined oil products and include only to a limited degree upstream operations such as drilling and production of crude oil, natural gas, and natural gas liquids. Upstream workers are exposed to benzene through contact with crude oil and natural gas, and some of these workers experience high levels of benzene exposure over shorter periods of time [18, 19]. In addition, benzene has been reported to be found as a laboratory chemical (solvent), as a component or a degradation product in drilling mud, as a minor component in diesel exhaust, and possibly as a degreasing agent [20, 21]. A marked limitation of the earlier studies in the petroleum industry is the likely presence of a healthy worker effect due to an overall mortality and overall cancer incidence among these workers that are significantly lower than in the general population used as a reference [12, 13, 22]. The healthy worker effect might mask increased risks 123 14 Cancer Causes Control (2008) 19:1323 of both multiple myeloma and non-Hodgkin lymphoma even in studies capable of showing an increased risk of acute myelogenous leukemia. We studied whether workers employed in Norway's upstream offshore petroleum industry have a higher risk of developing hematologic neoplasms than the general working population in Norway. We focused on the differences in risk of leukemia subtypes, multiple myeloma and non-Hodgkin lymphoma in the cohort combined and among different job categories. Materials and methods Study population and study design We performed a historical cohort study of the cancer incidence in Norway's upstream petroleum industry. Statistics Norway established the cohort using the information from the Norwegian Registry of Employers and Employees, which is owned by the Norwegian Labour and Welfare Organisation. All employers are required to register their employees with a personal identification number, industrial classification code (International Standard Industrial Classification (ISIC) or the Classification of Economic Activities in the European Union (NACE)), county of work, and the first and last date of all their engagements. The Registry requires that all engagements with a mean of 4 h of work per week, provided that the engagement lasts for at least 6 weeks, must be registered. The Registry included on July 31, 2004, a total of 1,961,711 workers contributing with 2,126,699 work engagements [23]. The Norwegian Registry of Employers and Employees was established in 1978, became operational in 1983 and contains engagements from 1981 and onwards. Since the Cancer Registry, at time of study, only included cases until 2003, we used engagements from 1981 until 2003. Norway's petroleum industry has been operating offshore since the early 1970s and thus the cohort does not include all workers who were engaged during the period 19701980. Still, many of these workers might be included in the cohort with possible new engagements registered after 1981. This means that some of these early subjects might have had a longer engagement offshore than registered in our cohort. The criteria we used for the cohort of petroleum workers were workers registered with one of the following offshore-related industrial classification codes: ISIC 22 (extraction of crude oil and natural gas), ISIC 5032 (oil drilling), NACE 11100 (extraction of crude oil and natural gas), and NACE 11200 (service activities incidental to oil and gas extraction excluding surveying), or having Norway's continental shelf (North Sea) as the work location. Based on the workers' location of work (onshore or offshore) and the industrial classification codes for their first registered engagement in the offshore-related petroleum industry, we categorized the petroleum workers into five job categories: (1) upstream operator offshore, (2) drilling and well maintenance offshore, (3) catering offshore, (4) others offshore and (5) petroleum workers onshore. The category ``upstream operator offshore'' contained solely workers registered with the NACE and ISIC code ``extraction of crude oil and natural gas''. These workers mainly work in the production and processing unit. This includes job categories such as process technicians, laboratory engineers, control operators, and other job groups involved in the production process including stabilization, separation, and fractionation of the crude oil, natural gas, and natural gas liquids. The category ``drilling and well maintenance offshore'' includes the ISIC code 50230 (oil drilling) and NACE code 11200 (service activities incidental to oil and gas extraction excluding surveying). NACE code 11200 comprises activities such as drilling of wells and installation, disassembling, and maintenance of drilling towers at site on contract and includes job groups such as drill floor crew, derrick employees, mud loggers and engineers, shale shaker operators, and well service crew. The category ``catering offshore'' includes job groups such as catering crew, chefs and housekeeping personnel. The category ``others offshore'' includes miscellaneous industrial codes and comprises activities contracted out to oil field service companies, such as construction and maintenance personnel and logistics. Finally, ``petroleum workers onshore'' contains workers registered with an offshore-related engagement without being registered with the North Sea as the location of work. This job category contains workers involved in administering, planning, and coordinating the activities offshore. We drew up to six referents per petroleum worker at random from the general working population, using the same Norwegian Registry of Employers and Employees and the same year of the first engagement of the corresponding petroleum worker. We matched the referents to the petroleum worker by gender, age, and community of residence. The crude historical cohort included 71,018 workers from the petroleum industry (``at risk'') and 424,584 referents. We excluded subjects from the cohort if they had a cancer diagnosis before entering into the cohort, and referents if they had an earlier engagement in the petroleum industry before they were drawn as referents even if they were not considered exposed in the engagement. We allowed subjects to serve as referents for more than one ``subject at risk''. The final cohort included 27,919 offshore workers (89% men) distributed on the four 123 Cancer Causes Control (2008) 19:1323 Table 1 Characteristics of the study population studied to determine whether workers employed in Norway's upstream offshore petroleum industry have a higher risk of developing hematologic neoplasms than the general working population in Norway, 19812003 Variable Referents Petroleum workers (all) Onshore workers (all) Offshore workers Upstream operator offshore Drilling and well maintenance Catering offshore Others offshore Number of subjects Male (%) Female (%) Age, mean (standard deviation) At inclusion into the cohort At the end of follow-up Mean education level (16) Tertiary education (%) Intermediate education (%) Compulsory education (%) Start date in industry (%) Before 1986 19861990 19912003 Average follow up (years) Person-years at follow-up 366,114 283,002 (77.3) 83,112 (22.7) 33.8 (9.5) 46.3 (11.5) 4.2 (1.5) 28.4 59.0 11.4 28.3 19.4 52.3 11.5 4,213,716 70,600 55,376 (78.4) 15,224 (21.6) 33.8 (9.5) 46.4 (11.4) 4.7 (1.6) 34.4 53.8 6.4 28.3 19.3 52.3 11.5 815,049 42,681 30,611 (71.7) 12,070 (28.3) 33.9 (9.7) 46.3 (11.5) 5.1 (1.6) 46.8 41.7 4.3 29.0 16.1 54.9 11.3 482,987 6,734 5,853 (86.9) 881 (13.1) 33.8 (8.1) 49.6 (11.3) 4.2 (1.2) 18.4 72.5 7.2 49.1 22.8 28.1 14.6 98,341 7,049 6,722 (95.4) 327 (4.6) 32.6 (8.6) 43.4 (11.2) 4.0 (1.2) 13.9 74.4 9.6 15.0 17.2 67.8 9.7 68,238 2,417 1,231 (50.9) 1,186 (49.1) 32.6 (10.0) 47.9 (11.4) 3.5 (1.1) 8.7 73.6 16.4 47.7 19.9 32.4 13.9 33,677 11,719 10,959 (93.5) 760 (6.5) 34.3 (9.6) 46.7 (10.7) 4.0 (1.3) 15.7 70.8 9.8 18.0 30.3 51.7 11.2 131,808 123 15 16 Cancer Causes Control (2008) 19:1323 offshore job categories comprising 332,064 person-years (Table 1). Statistics Norway established and linked the total cohort to the Cancer Registry of Norway in April 2006, including all cases of cancer reported up to December 31, 2003 with information on the date of diagnosis and the diagnosis (location, morphology, and histology). For the localization of the neoplasms, the cancer cases were coded according to a modified version of the International Classification of Diseases (ICD-7, three-digit codes). For subjects with more than one cancer, only the first was used to identify hematological neoplasms. The Cancer Registry of Norway is based on reporting from multiple sources, such as physicians, pathology laboratories, and death certificates from Statistics Norway mentioning cancer or cancer-related illnesses, ensuring a high degree of accuracy and completeness. Statistics Norway also linked the cohort to the Norwegian Cause of Death Registry and the Norwegian Education Registry, including the variable highest completed education, ranging from 1 (elementary school) to 6 (PhD degree). Statistical analysis We estimated the rate ratios comparing the various working categories with the general working population using the Cox proportional hazard regression model. We censored subjects at the end of follow-up (December 31, 2003), the date of death or date of diagnosis of another type of cancer than the one being studied, whichever occurred first. We checked the proportional hazards assumption for overall cancer and all hematologic neoplasms by comparing the estimated--lnln survivor curves for the different groups being investigated. There was no marked deviation from the proportional hazards assumption. We performed multivariate analysis including the independent covariates age, gender, year of first registered engagement and educational level. Age was defined at time of entering into the cohort (time of first registered engagement) and used as a continuous variable in the model. To ensure a sufficient adjustment for the year of first engagement, we repeated all analyses of the ``upstream operators offshore'' using only the referents drawn for this specific job category. We also repeated the analysis excluding the upstream operators offshore being registered only 1 year or less in the offshore related industry offshore. Finally, we also performed subanalysis in 5 years periods according to the year of first registered engagement. We performed all analysis using SPSS 14.0.1 (SPSS Inc., Chicago, IL, USA). Ethical considerations We conducted the study with the approval of the Western Norway Regional Committee for Medical Research Ethics, the Norwegian Data Inspectorate and the Directorate for Health and Social Affairs. Rikstrygdeverket (now merged into the Norwegian Labour and Welfare Organisation) gave us permission to use the Norwegian Registry of Employers and Employees, the Cancer Registry of Norway provided data, and Statistics Norway (Ministry of Finance) approved the use of information from the Norwegian Cause of Death Registry and the Norwegian Education Registry. Results The incidence of overall cancer (all sites) among the offshore workers did not differ significantly from that of the general working population in any job category (Table 2). However, the overall mortality was slightly increased among the offshore workers, ascribed to an increased mortality from all causes in the job categories ``catering personnel offshore'' (rate ratio (RR) 1.44) and ``other personnel offshore'' (RR 1.30). The mortality ratio was below unity for the job category ``upstream operators offshore'' with borderline significance (RR 0.87). ``Upstream operators offshore'' had an excess risk of hematologic neoplasms (RR 1.90) (Table 2). The increased risk among ``upstream operators offshore'' was ascribed to an increased risk of acute myelogenous leukemia (RR 2.89) and multiple myeloma (RR 2.49) (Table 3). The other job categories had no increased risk of either hematologic neoplasms combined (blood and bone marrow), acute myelogenous leukemia or multiple myeloma. We also compared ``upstream operators offshore'' to a reference group comprising only referents drawn for this specific job category during the establishment of the cohort (n = 35,634). The risk increased slightly in this subanalysis for both hematologic neoplasms localized to blood and bone marrow (RR 2.00, 95% CI 1.18.3.38), acute myelogenous leukemia (RR 3.67, 95% CI 1.3010.34), and multiple myeloma (RR 2.54, 95% CI 1.095.90). The mortality from all causes and incidence of cancer (all sites) for ``upstream operators offshore'' did not differ from that of this alternative reference group. Further, by excluding the upstream operators offshore being registered only for 1 year or less in the offshore-related industry offshore (n = 71), no change in the risk estimates was found. The risk of lymphoma for the total group of offshore workers was significantly lower than in the general working population (RR 0.64) (Table 2). However, no significant difference was found within any of the job categories 123 Table 2 Rate ratios (RR) for overall mortality (all causes of death), cancer incidence (all sites of cancer), hematologic neoplasms localized to blood and bone marrow and hematologic neoplasms localized to lymphatic tissues (lymphomas) for workers in the job categories in the upstream petroleum industry versus referents from the general working population, Norway 1981 2003 adjusted for gender, age at inclusion into the cohort, year of first exposure, and education using the Cox proportional hazards model Cancer Causes Control (2008) 19:1323 All causes of death All sites of cancer Hematologic neoplasms (blood and bone marrow) Lymphomas n RR 95% CI n RR 95% CI n RR 95% CI n RR 95% CI Referents 10,002 11,271 394 655 All petroleum workers offshore 898 1.13 1.061.21 815 1.01 0.941.09 37 1.24 0.881.74 33 0.64 0.450.91 Upstream operators offshore 227 0.87 0.761.00 251 0.93 0.821.06 20 1.90 1.193.02 14 0.82 0.471.42 Drilling and well maintenance 160 1.08 0.921.27 150 0.96 0.821.13 9 1.55 0.803.02 7 0.66 0.311.40 Catering offshore 123 1.44 1.201.73 102 1.04 0.851.27 2 0.68 0.172.73 1 0.20 0.031.41 Others offshore 388 1.30 1.171.44 312 1.10 0.981.23 6 0.58 0.261.30 11 0.60 0.331.08 Rate ratios significantly different from unity are given in bold. The letter n denotes number of cases 17 offshore or for any of the subtypes of lymphomas (Table 3). In the subanalyses in 5-year periods according to the first registered engagement we found strong effects for the first 5-year period 19811985 for both acute myelogenous leukemia (RR 3.26), and multiple myeloma (RR 2.85). The RRs for having the first registered engagement prior to 1986 and in the period 19862003, respectively, are given in Table 4 for all offshore workers combined and the job category ``upstream operators offshore''. Among the ``upstream operators offshore'', 29 of the total of 34 cases of hematologic neoplasms occurred among the workers participating in the first 5-year period. Compared with the general working population, the overall mortality (all causes) and overall cancer (all sites) was relatively stable for these periods both among all offshore workers combined and for ``upstream operators offshore''. Discussion Offshore petroleum workers employed on Norway's continental shelf from 1981 to 2003 had a higher risk of developing hematologic neoplasms, especially multiple myeloma and acute myelogenous leukemia, than the general working population. The risk was elevated among ``upstream operators offshore''. These are the workers assumed to have had the most extensive contact with different phases of crude oil and other products containing benzene. No increased risk of developing lymphoma in this group of workers was found. Given the established association between benzene exposure and hematologic neoplasms [1, 2], benzene exposure probably caused the observed increased risk of acute myelogenous leukemia and multiple myeloma. At present, it is unclear at which level of exposure benzene poses an increased risk of developing hematologic neoplasms [9, 17, 2427]. Previous studies that assessed the association between benzene exposure and hematologic neoplasms in cohorts of petroleum workers were limited by the lack of good exposure estimates including information on the variability of benzene exposure. This is also a major limitation of our study. Information about past exposure to benzene through contact with petroleum streams in Norway's offshore petroleum industry is scarce [28], and we do not have any objective data on a possible additional exposure through the use of benzene as a degreasing agent, as a laboratory chemical or as a component in drilling mud. However, crude oil assays from different regions of the Norwegian continental shelf (n = 14) reported a mean and median value of 0.28% benzene by weight, within a range of \0.010.66% [19]. Pooled full shift measurements from processing and drilling areas in the period 19942003 123 18 Cancer Causes Control (2008) 19:1323 Table 3 Rate ratios (RR) and 95% confidence intervals (95% CI) for hematologic neoplasms among workers in job category ``upstream operators offshore'' versus referents from the general working population, Norway 19812003 Referents Upstream operators offshore n n RR 95% CI All hematologic neoplasms localized to blood and bone marrow Multiple myeloma Acute lymphocytic leukemia Chronic lymphocytic leukemia Acute myelogenous leukemia Chronic myelogenous leukemia Other hematologic neoplasms All lymphomas Non-Hodgkin lymphoma Hodgkin disease Unspecified and undefined lymphomas 393 132 17 73 86 31 54 655 524 114 17 20 1.90 1.193.02 9 2.49 1.215.13 1 2.17 0.2916.6 3 1.62 0.515.20 6 2.89 1.256.67 1 1.44 0.1910.7 0 14 0.82 0.471.42 14 1.01 0.581.75 0 0 Rate ratios are adjusted for gender, age at inclusion into the cohort, year of first exposure, and education using the Cox proportional hazard regression model. Rate ratios significantly different from unity are given in bold. The letter n denotes number of cases Table 4 Rate ratios (RR) and 95% confidence intervals (95% CI) for different types of hematologic neoplasms, all cancer sites and overall mortality (all causes of death) for all offshore workers combined and workers in the job category ``upstream operators offshore'' versus referents from the general working population Referents Offshore workers combined Upstream operators offshore n n RR 95% CI n RR 95% CI First exposure 19811985 All hematological neoplasms All leukemia Acute myelogenous leukemia Multiple myeloma Non-Hodgkin lymphoma All cancers All causes of death 549 107 45 83 271 6,075 5,884 45 1.08 0.791.48 29 1.38 0.932.03 14 1.84 1.053.22 10 2.49 1.304.78 7 2.24 1.005.01 5 3.26 1.288.30 11 1.74 0.903.38 9 2.85 1.375.93 19 0.93 0.571.50 10 0.96 0.491.86 411 1.00 0.901.11 187 0.91 0.781.06 487 1.13 1.021.24 184 0.89 0.761.03 First exposure 19862003 All hematological neoplasms All leukemia Acute myelogenous leukemia Multiple myeloma Non-Hodgkin lymphoma All cancers All causes of death 500 100 41 49 253 5,196 4,118 25 0.65 0.430.97 4 0.50 0.181.37 4 1.23 0.443.44 5 1.52 0.603.86 10 0.50 0.270.95 404 1.04 0.931.15 411 1.15 1.031.27 5 0.78 0.321.89 1 0.80 0.115.75 1 1.76 0.2412.86 0-- -- 4 1.22 0.453.29 64 0.97 0.761.25 43 0.79 0.591.08 Rate ratios are stratified on the year of the first engagement in the offshore-related petroleum industry (``first exposure'') and adjusted for gender, age at inclusion into the cohort, year of first exposure, and education using the Cox proportional hazard regression model. Rate ratios significantly different from unity are given in bold. The letter n denotes number of cases. (n = 367) show an arithmetic mean benzene exposure of 0.037 ppm, ranging from below level of detection to 2.6 ppm [28]. Recent studies report a benzene exposure averaged over a full shift well below 0.1 ppm during ordinary activity, ranging from below 0.001 to 0.69 ppm [19, 29]. Low full shift exposures for most job categories are also reported internationally from the upstream petroleum industry [20, 3032]. However, specific tasks causing high short-term exposures, such as cleaning of tanks containing residues of crude oil with a reported maximum of 123 Cancer Causes Control (2008) 19:1323 19 16.8 ppm benzene [19], are major determinants of exposure and may contribute significantly to the total benzene exposure [19, 29]. Collins et al. [33] reported that the number of peak exposures to benzene (above 100 ppm), rather than cumulative exposure, might be the best predictor of risk of hematologic neoplasms. Nevertheless, in spite of the lack of good exposure estimates for the groups included in our study, we assume that the ``upstream operators offshore'', who have had the most extensive contact with crude oil and natural gas, was highest and most homogenously exposed to benzene. The reported exposure levels, and the low content of benzene in the crude oil, still indicate that the atmospheric exposure from the crude or condensate has been well below the exposure levels that most studies have reported to be necessary for inducing hematologic neoplasms. This increased risk thus implies that the exposure levels have been higher than published for this industry, or that the increased risk for these neoplasms can be found at lower levels of exposure than previously assumed. In support of the latter interpretation a few studies have reported an association between a mean benzene exposure around, and even below, 1 ppm and increased risk of acute myelogenous leukemia [9], increased single-strand breaks in DNA of leukocytes [34, 35], chronic hematotoxic effects including reduced neutrophil, lymphocyte and platelet counts in peripheral blood [36], and acute alterations of the immune system including levels of circulating lymphocytes and immunoglobulins [37], respectively. The sensitivity to benzene's toxic effects differs between individuals, partly explained by genetic polymorphisms in enzymes involved in the metabolism of benzene and genes which are of importance in DNA repair pathway or the regulation of hematopoiesis [3842]. Hence, the present and previous observations suggest that benzene even at low exposure levels may contribute to the risk of hematologic neoplasm, especially in genetically susceptible individuals. As cigarette smoke is a known source of benzene exposure [43], a leukemogenic effect at low benzene-exposure is further supported by the reported increase in acute myelogenous leukemia and cytogenetic abnormalities in smokers [44]. Smokers not occupationally exposed to benzene reached a morning concentration of benzene in blood of up to 13 nmol/l after smoking four or five cigarettes [45], which is equivalent to a benzene exposure in the breathing zone of approximately 0.3 ppm averaged over an 8-h shift [46]. The association between benzene exposure and the development of specific subtypes of leukemia is still unclear. Our results showing the highest risk for acute myelogenous leukemia are in accordance with findings in previous studies [1, 2]. The risk of acute lymphocytic leukemia was also elevated but was based only on two cases in the exposed groups and therefore not statistically significant. Hence, the statistical power of our study to detect an increased risk for the more infrequent subtypes of leukemias might be too low. This is also reflected in the wide confidence intervals for these subtypes. The most recent meta-analysis of benzene exposure and leukemia subtypes including nine cohorts and 13 casecontrol studies from several industries, found a high and significant risk of acute myelogenous leukemia, with a positive dose response relationship across study designs [2]. The risk for developing chronic lymphocytic leukemia was increased in the casecontrol studies, but not in the cohort studies. The data for chronic myelogenous leukemia and acute lymphocytic leukemia were sparse and inconclusive. The association between benzene exposure and multiple myeloma is a contentious issue [3, 4, 47]. A meta-analysis of 22 cohort mortality studies consisting of 250,000 petroleum workers concluded that these workers do not have an increased risk of multiple myeloma as a result of their exposure to benzene, benzene-containing liquids or other petroleum products in their work environment [16]. In contrast, a more recent meta-analysis including seven cohort studies focusing on benzene-exposed workers reported a significant excess in the relative risk of multiple myeloma in relation to benzene exposure (relative risk 2.13, 95% CI: 1.313.46) [47]. The majority of the studies included in the negative meta-analysis [16] were performed on refinery and distribution workers, and it has been questioned whether the cohort members were sufficiently exposed to benzene to detect any effect [47]. Our finding of an increased risk of multiple myeloma among the upstream petroleum workers, also showing an increased risk of acute myelogenous leukemia, provides further evidence of an association between benzene exposure and the risk of multiple myeloma. The risk of acute myelogenous leukemia and multiple myeloma was highest among the workers who had their first registered engagement in the offshore petroleum industry at the beginning of the study period. A plausible explanation for this finding might be a general improvement of the working environment offshore, including reduced benzene exposure, following regulations and other initiatives in the 1990s. Another explanation might be that the follow-up time for workers starting later was too short to detect any increased risk. While maximum follow-up in the present study is truncated at 22 years for both offshore workers and referents, the average follow-up periods were 14.6 and 11.5, respectively. Several authors have discussed the temporal variation of the risk of developing leukemia after exposure to benzene. For the Pliofilm-cohort [27], the increased risk was reported to be attributable primarily to exposures occurring during the 10 years preceding the death of the case rather than the more distant exposure [48, 49], and 123 20 Cancer Causes Control (2008) 19:1323 that the risk was highest in the first few years after cessation of exposure [49]. Glass et al. [24] have reported a similar pattern prior to diagnosis of leukemia in the Health Watch cohort from the Australian petroleum industry. Further, in the large cohort of benzene-exposed Chinese workers the relative risk was increased even for workers exposed to benzene less than 5 years [25]. On the other hand, mortality studies of workers from the petroleum industry have shown an increased risk of acute myelogenous leukemia along with increasing duration of employment, with the highest risk among workers employed for more than 20 30 years [10, 11, 50]. Although the time estimates reported in these studies represent a combination of latency and the effect of cumulative exposure, these observations are compatible with a wide range of latency periods for acute myelogenous leukemia induced by benzene, and indicate that an increased risk might be found also for work in the later periods in the present study after longer follow-up period. Previous studies report that the time from first exposure till diagnosis or death of multiple myeloma possibly related to benzene is above 20 years [27, 33, 47]. In our study, the median time between the first year of registered engagement and diagnosis of acute myelogenous leukemia and multiple myeloma for cases in the job category ``upstream operators offshore'' included into the cohort prior to 1986 was 6 (range 521) and 18 (range 1320) years, respectively. This is compatible with a longer latency in multiple myeloma, but the rather short follow-up time together with the low number of cases, limit the possibility of a powerful discussion on the latency issue based on our data. The etiology of lymphomas is still largely unknown, but given the potential for benzene to affect the immune system an association between benzene exposure and particularly non-Hodgkin lymphoma has been suggested [7, 8, 25]. We found that the total cohort of offshore workers had a decreased risk of lymphomas. However, there were no statistical difference in the risk of neither non-Hodgkin lymphoma nor Hodgkin disease in any of the job categories. This is in line with what has been found in most other cohorts of petroleum workers [15, 17]. The increased risk of both acute myelogenous leukemia and multiple myeloma indicates a sufficient exposure to benzene in this cohort to induce hematologic neoplasms, and the negative findings on lymphomas therefore argue against an association between benzene and lymphomas at these exposure levels. However, in the large cohort of Chinese workers [25] it was reported a relative risk of non-Hodgkin lymphoma of 4.2 for workers with 10 or more years of benzene exposure, as compared to a relative risk of 0.7 for workers with exposure less than 5 years. The increased risk was found only for workers with an average exposure of 25 ppm benzene (RR 4.7). Thus, in our study the follow-up time to detect an increased risk of non-Hodgkin lymphoma might also have been too short. Smoking is associated with risk of acute myelogenous leukemia [44] and might also be associated with multiple myeloma [51]. We do not have data on smoking in our study. However, the risk estimates were adjusted for the level of education, which is sometimes used as a surrogate measure of social class and smoking [52, 53]. The ``upstream operators offshore'' had a modest and non-significantly increased risk of cancer in lung and bronchus, while a slightly higher increased risk was found for the two groups ``catering offshore'' and ``others offshore'' where no increased risk of hematologic neoplasms was found (data not shown). Further, the percentages of smokers among operators in offshore oil industry in Norway in 1997 was reported to be 31% [54], while the percentage of smokers in the general male working population in Norway was estimated to be 35% in 19971999 [55]. All these factors indicate that smoking did not contribute substantially to the increased risk of hematologic neoplasms among ``upstream operators offshore''. Given the complexity of the exposure to crude oil and to other agents at offshore installations, other factors might have contributed to the increased risk of hematologic neoplasms. Ionizing radiation has been shown to cause both acute myelogenous leukemia [56] and multiple myeloma [57]. Some petroleum workers are potentially exposed to ionizing radiation due to nondestructive testing of welding seams, well-logging, and contact with sediments emitting low-level ionizing radiation. The majority of exposure measurements relevant for the petroleum industry offshore have been reported to be well below the present recommended annual limit values of 1 mSv, and all below that of 20 mSv for occupational radiation exposure [58, 59]. From naturally occurring radioactive materials in the oil and gas industry, maximal effective doses ranging from 0.04 to 0.60 mSv for normal activities and from 0.08 to 1.20 mSv during revision stops have been estimated [60]. The healthy worker effect is a bias that might occur in occupational mortality studies comparing the working population of interest with general populations, possibly masking an increased risk of the disease under study. This has been shown for mortality risks for both chronic diseases and cancer [61], and is a potential limitation of previous studies in the petroleum industry reporting a significantly lower overall mortality of causes and overall cancer incidence among these workers than in the general population [1113, 22, 50]. In addition to using a historical prospective design, we drew our referents from the general working population and from the same 123 Cancer Causes Control (2008) 19:1323 21 registry as the subjects ``at risk'', hence reducing this possible healthy worker effect. We also matched the subjects and referents for level of education and place of residence, thereby reducing a potential bias due to culture, ethnicity, lifestyle and/or regional environmental pollution, as discussed by Leonard et al. [61]. The Cancer Registry of Norway was the source for cases of cancer among both petroleum workers and referents, thereby avoiding classification bias of the outcome. The incidence of overall cancer (all sites) among the offshore workers did not differ significantly from that of the general working population in any job category, indicating that our study design has sufficiently reduced such possible information and selection biases. However, the overall mortality was slightly increased among the offshore workers compared with the general working population (RR 1.13), ascribed to a significantly increased mortality from all causes in the job categories ``catering offshore'' (RR 1.44) and ``others offshore'' (RR 1.30). The mortality ratio was below unity for the job category ``upstream operators offshore'' (RR 0.87) with borderline significance, indicating that a healthy worker effect still might have been present in this specific work category. In conclusion, workers in the upstream petroleum industry exposed to crude oil and other hydrocarbons had a significantly increased risk of developing acute myelogenous leukemia and multiple myeloma, but not nonHodgkin lymphoma. The workers employed early in the study period had the highest risk. Although we cannot exclude a possible contribution of other types of specific or combined exposure, occupational exposure to benzene is the most likely candidate for the increased risk. However, given the lack of detailed information on occupation, job tasks, and exposure in the groups studied, exposure was probably nondifferentially misclassified in this historical prospective study. Such nonspecific exposure indicators might have caused an underestimation of the risk for the individuals with substantial exposure. A casecontrol study nested within the cohort should be performed to improve the estimates of the association between the various types of exposure and the risk of the subtypes of hematologic neoplasms. Acknowledgments We acknowledge Trond Pedersen and coworkers of Statistics Norway for establishing the historical cohort, Tom Brge Johannesen and Tove Dahl of the Cancer Registry of Norway for helping us with categorization of the diagnose of the various hematological neoplasms, and A ge Johannesen, also of the Cancer Registry of Norway, for preparing the Registry for matching the cohort to the Registry. This project has been financed with the aid of the Research Council of Norway, EXTRA funds from the Norwegian Foundation for Health and Rehabilitation, and the Department of Health of UNIFOB AS. We declare that we have no conflict of interest. References 1. Benzene. IARC Monographs on the evaluation of carcinogenic risks to humans, vol 29 (Suppl 17). International Agency for Research on Cancer, Lyon, France, 1987 2. Schnatter AR, Rosamilia K, Wojcik NC (2005) Review of the literature on benzene exposure and leukemia subtypes. Chem Biol Interact 153154:921 3. Bergsagel DE, Wong O, Bergsagel PL et al (1999) Benzene and multiple myeloma: appraisal of the scientific evidence. Blood 94(4):11741182 4. Goldstein BD, Shalat SL (2000) The causal relation between benzene exposure and multiple myeloma [letter to editor]. Blood 95(4):15121514 5. Wong O, Raabe GK (2000) Non-Hodgkin's lymphoma and exposure to benzene in a multinational cohort of more than 308,000 petroleum workers, 1937 to 1996. J Occup Env Med 42(5):554568 6. Goldstein BD, Shalat S (2000) Non-Hodgkin's lymphoma and exposure to benzene in petroleum workers [letter to editor]. J Occup Env Med 42(12):11331134 7. Mehlman MA (2006) Causal relationship between non-Hodgkin's lymphoma and exposure to benzene and benzene-containing solvents. Ann N Y Acad Sci 1076:120128 8. Smith MT, Jones RM, Smith AH (2007) Benzene exposure and risk of non-Hodgkin lymphoma. Cancer Epidemiol Biomarkers Prev 16(3):385391 9. Glass DC, Gray CN, Jolley DJ et al (2003) Leukemia risk associated with low-level benzene exposure. Epidemiology 14:569577 10. Sathiakumar N, Delzell E, Cole P, Brill I, Frisch J, Spivey G (1995) A casecontrol study of leukemia among petroleum workers. J Occup Environ Med 37(11):12691277 11. Divine BJ, Hartman CM (2000) Update of a study of crude oil production workers 194694. Occup Environ Med 57(6):411417 12. Gun RT, Pratt NL, Griffith EC, Adams GG, Bisby JA, Robinson KL (2004) Update of a prospective study of mortality and cancer incidence in the Australian petroleum industry. Occup Environ Med 61(2):150156 13. Satin KP, Bailey WJ, Newton KL, Ross AY, Wong O (2002) Updated epidemiological study of workers at two California petroleum refineries, 195095. Occup Environ Med 59(4):248 256 14. Thomas TL, Waxweiler RJ, Moure-Eraso R, Itaya S, Fraumeni JF Jr (1982) Mortality patterns among workers in three Texas oil refineries. J Occup Med 24(2):135141 15. Goldstein BD (1990) Is exposure to benzene a cause of human multiple myeloma? Ann N Y Acad Sci 609:225230 16. Wong O, Raabe GK (1997) Multiple myeloma and benzene exposure in a multinational cohort of more than 250,000 petroleum workers. Regul Toxicol 26:188199 17. Schnatter AR, Armstrong TW, Nicolich MJ et al (1996) Lymphohaematopoietic malignancies and quantitative estimates of exposure to benzene in Canadian petroleum distribution workers. Occup Environ Med 53(11):773781 18. Durand KTH, Lees PSJ, Kern DG (1995) Exposure assessment and respirator selection in the cleaning of crude oil process vessels. Appl Occup Environ Hyg 10(2):120124 19. Kirkeleit J, Riise T, Bratveit M, Moen BE (2006) Benzene exposure on a crude oil production vessel. Ann Occup Hyg 50:123129 20. Whiteley S, Plant N (2000) Occupational benzene, toluene, xylene and ethylbenzene during routine offshore oil and gas production operations. HSE offshore technology report OTO 1999 088. UK Health and Safety Executive, Bootle, UK. 123 22 Cancer Causes Control (2008) 19:1323 Available from: http://www.hse.gov.uk/research/otopdf/1999/ oto99088.pdf. Accessed 14 August, 2007 21. Gardner R (2003) Overview and characteristics of some occupational exposures and health risks on offshore oil and gas installations. Ann Occup Hyg 47(3):201210 22. Sorahan T, Nichols L, Harrington JM (2007) Mortality of UK oil refinery and petroleum distribution workers, 19512003. Occup Med 57(3):177185 23. Mehlum IS, Kjuus H (2005) Omfang og konsekvenser av arbeidskader og arbeidsbetinget sykdom pa norsk kontinentalsokkel [The extent and consequences of work related injuries and diseases on Norways continental shelf] (in Norwegian). Natl Inst Occup Health 4:918. Available at: http://www.stami.no/./ ?module=Articles;action=Article.publicShow;ID=2276. Accessed August 15, 2007 24. Glass DC, Sim MR, Fritschi L, Gray CN, Jolley DJ, Gibbons C (2004) Leukemia risk and relevant benzene exposure period--re: follow-up time on risk estimates, Am J Ind Med 42:481489, 2002. Am J Ind Med 45(2):222223 25. Hayes RB, Yin SN, Dosemeci M et al (1997) Benzene and the dose-related incidence of hematologic neoplasms in China. Chinese Academy of Preventive Medicine--National Cancer Institute Benzene Study Group. J Natl Cancer Inst 89(14):1065 1071 26. Hayes RB, Songnian Y, Dosemeci M, Linet M (2001) Benzene and lymphohematopoietic malignancies in humans. Am J Ind Med 40(2):117126 27. Rinsky RA, Hornung RW, Silver SR, Tseng CY (2002) Benzene exposure and hematopoietic mortality: a long-term epidemiologic risk assessment. Am J Ind Med 42(6):474480 28. Steinsvag K, Bratveit M, Moen BE (2007) Exposure to carcinogens for defined job categories in Norway's offshore petroleum industry, 1970 to 2005. Occup Environ Med 64(4):250258 29. Bratveit M, Kirkeleit J, Moen BE (2007) Biological monitoring of benzene exposure for process operators during ordinary activity in the upstream petroleum industry. Ann Occup Hyg. Epub ahead of print: doi:10.1093/annhyg/mem029 30. Runion HE (1988) Occupational exposure to potentially hazardous agents in the petroleum industry. Occup Med 3:431444 31. Glass DC, Adams GG, Manuell RW, Bisby JA (2000) Retrospective exposure assessment for benzene in the Australian petroleum industry. Ann Occup Hyg 44(4):301320 32. Verma DK, Johnson DM, McLean JD (2000) Benzene and total hydrocarbon exposures in the upstream petroleum oil and gas industry. AIHAJ 61:255263 33. Collins JJ, Ireland B, Buckley CF, Shepperly D (2003) Lymphohaematopoeitic cancer mortality among workers with benzene exposure. Occup Environ Med. 60(9):676679 34. Nilsson R, Nordlinder R, Hgstedt B, Karlsson A, Jarvholm B (1996) Genotoxic effects in workers exposed to low levels of benzene from gasoline. Am J Ind Med 30(3):317324 35. Sul D, Lee E, Lee MY et al (2005) DNA damage in lymphocytes of benzene exposed workers correlates with trans,trans-muconic acids and breath benzene levels. Mutat Res 582(12):6170 36. Lan Q, Zhang L, Li G et al (2004) Hematotoxicity in workers exposed to low levels of benzene. Science 306(5702):17741776 37. Kirkeleit J, Ulvestad E, Riise T, Bratveit M, Moen BE (2006) Acute suppression of serum IgM and IgA in tank workers exposed to benzene. Scand J Immunol 64:690698 38. Rothman N, Smith MT, Hayes RB et al (1997) Benzene poisoning, a risk factor for hematological malignancy, is associated with the NQO1 609C?T mutation and rapid fractional excretion of chlorzoxazone. Cancer Res 57(14):28392842 39. Wan J, Shi J, Hui L et al (2002) Association of genetic polymorphisms in CYP2E1, MPO, NQO1, GSTM1, and GSTT1 genes with benzene poisoning. Environ Health Perspect 110(12):12131218 40. Shen M, Lan Q, Zhang L et al (2006) Polymorphisms in genes involved in DNA double-strand break repair pathway and susceptibility to benzene-induced hematotoxicity. Carcinogenesis 27(10):20832089 41. Lan Q, Zhang L, Shen M et al (2005) Polymorphisms in cytokine and cellular adhesion molecule genes and susceptibility to hematotoxicity among workers exposed to benzene. Cancer Res 65(20):95749581 42. Lv L, Kerzic P, Lin G et al (2007) The TNF-alpha 238A polymorphism is associated with susceptibility to persistent bone marrow dysplasia following chronic exposure to benzene. Leuk Res. Epub ahead of print: doi:10.1016/j.leukres.2007.01.014 43. Darrall KG, Figgins JA, Brown RD, Phillips GF (1998) Determination of benzene and associated volatile compounds in mainstream cigarette smoke. Analyst 123:10951101 44. Lichtman MA (2007) Cigarette smoking, cytogenetic abnormalities, and acute myelogenous leukemia. Leukemia 21(6):1137 1140 45. Pekari K, Vainiotalo S, Heikkila P, Palotie A, Luotamo M, Riihimaki V (1992) Biological monitoring of occupational exposure to low levels of benzene. Scand J Work Environ Health 18:317322 46. DFG (Deutsche Forschungsgemeinschaft) (2005) MAK- und BAT-Werte-Liste. Maximale Arbeitsplatzkonzentrationen und Biologische Arbeitsstoff-toleranzwerte. Wiley-VCH, Weinheim 47. Infante PF (2006) Benzene exposure and multiple myeloma: a detailed meta-analysis of benzene cohort studies. Ann N Y Acad Sci 1076:90109 48. Finkelstein MM (2000) Leukemia after exposure to benzene: temporal trends and implications for standards. Am J Ind Med 38(1):17 49. Silver SR, Rinsky RA, Cooper SP, Hornung RW, Lai D (2002) Effect of follow-up time on risk estimates: a longitudinal examination of the relative risks of leukemia and multiple myeloma in a rubber hydrochloride cohort. Am J Ind Med 42:481489 50. Huebner WW, Wojcik NC, Rosamilia K, Jorgensen G, Milano CA (2004) Mortality updates (19701997) of two refinery/petrochemical plant cohorts at Baton Rouge, Louisiana, and Baytown, Texas. J Occup Environ Med 46(12):12291245 51. Nieters A, Deeg E, Becker N (2006) Tobacco and alcohol consumption and risk of lymphoma: results of a population-based casecontrol study in Germany. Int J Cancer 118(2):422430 52. Brnnum-Hansen J, Juel K (2004) Impact of smoking on the social gradient in health expectancy in Denmark. J Epidemiol Community Health 58(7):604610 53. Jha P, Peto R, Zatonski W, Boreham J, Jarvis MJ, Lopez AD (2006) Social inequalities in male mortality, and in male mortality from smoking: indirect estimation from national death rates in England and Wales, Poland, and North America. Lancet 368(9533):367370 54. Bull N, Riise T, Moen BE (1999) Influence of paternal exposure to oil and oil products on time to pregnancy and spontaneous abortions. Occup Med 49(6):371376 55. Riise T, Moen BE, Nordtvedt MW (2003) Occupation, lifestyle factors and health related quality of life: the Hordaland study. J Occup Environ Med 45:324332 56. Ionizing radiation (2000) Part 1: X- and gamma (c)-radiation, and neutrons. Summary of data reported and evaluation. IARC Monographs on the evaluation of carcinogenic risks to humans, vol 75. International Agency for Research on Cancer, Lyon, France 57. de Roos AJ, Baris D, Weiss NS et al (2006) Multiple myeloma. In: Schottenfeld D, Fraumeni JF Jr (eds) Cancer. Epidemiology 123 Cancer Causes Control (2008) 19:1323 23 and Prevention, 3rd edn. Oxford University Press, New York, pp 919945 58. Sekse T, Paulsen GU, Hannevik M et al (2005) Yrkeseksponering i Norge Ioniserende straling Ikke-ioniserende straling [Radiation exposure of workers in Norway]. Stralevern Rapport 2005:15. Norwegian Radiation Protection Authority, steras (in Norwegian) 59. Sources and effects of ionizing radiation: Volume 1: Sources. United Nations Scientific Committee on the Effects of Atomic Radiations. UNSCEAR 2000 report to the general assembly, with scientific annexes. United Nations, New York, 2000 60. Hamlat MS, Djeffal S, Kadi H (2001) Assessment of radiation exposures from naturally occurring radioactive materials in the oil and gas industry. Appl Radiat Isot 55(1):141146 61. Leonard RC, Kreckmann KH, Lineker GA, Mars G, Buchamich J, Youk A (2007) Comparison of standardized mortality ratios (SMRs) obtained from use of reference populations based on a company-wide registry cohort to SMRs calculated against local and national rates. Chem Biol Interact 166:317322 123