Benzene Exposure and Risk of Non-Hodgkin Lymphoma

Benzene is an important industrial chemical, as evidenced by production in excess of 2 billion gallons annually in the United States, and is a component of gasoline. Workers in a number of industries, including petroleum refining, shipping, rubber manufacturing, automobile repair, and shoe manufacturing, are potentially exposed to high levels of benzene. Benzene exposures among the general public are lower than most occupational exposures and result from cigarette smoking, gasoline use, and automobile emissions. Although benzene is a widely recognized cause of leukemia, the association of benzene with non-Hodgkin lymphoma (NHL) is less clear. Given the prevalence of benzene exposure and the potential for benzene to produce chromosome changes and other genetic changes of importance in NHL induction, it is essential to clarify the association between benzene exposure and NHL. Accurate quantification of the burden of benzene-associated diseases is necessary to evaluate the health risk posed by environmental contamination and occupational exposures at current exposure limits. Toward this aim, we undertook a review of all case-control and cohort studies that identified probable occupational exposures to benzene and NHL morbidity or mortality and concluded that the evidence supports an association between occupational benzene exposure and NHL.

In the early 1900s, benzene was widely used as an industrial solvent and through the appearance of multiple reports in the literature of anemia, pancytopenia, and leukemia, its toxicity to the blood and bone marrow quickly became manifest. French investigators were the first to report a case of lymphoma associated with benzene exposure in 1947 (1). Similar reports emerged in the 1960s from France, Italy, and Spain. Studies in Turkish shoe workers by Aksoy and others in the 1970s and 1980s were instrumental in establishing the relationship between benzene and leukemia, but these studies also identified 11 cases of lymphoma associated with occupational benzene exposure (2). They concluded that benzene could cause a variety of hematologic and lymphatic malignancies, in addition to damaging the bone marrow.

As highlighted in this issue by Vineis and coworkers, occupational exposure to solvents has been associated with an increased risk of NHL in numerous studies. In their review on the topic, Rego et al. (3) considered that 72% of studies with an accurate definition of solvent exposure observed positive associations between solvent exposures and NHL. Solvents, however, are chemically diverse and often exist as chemical mixtures, which makes it difficult to incriminate a specific chemical agent (4). Given the association of benzene with disorders of the bone marrow and blood, and the ability of the chemical to produce chromosomal and genetic changes important to NHL induction, it is plausible that benzene is a specific solvent associated with NHL.

We identified 43 case-control studies of NHL outcomes that recognized persons with probable occupational exposure to benzene. Forty of these 43 (93%) studies show some elevation of NHL risk, with 23 of 43 (53%) studies finding statistically significant associations between NHL risk and probable benzene exposure (Supplementary Table S1). Slightly more than half of these studies, 22 in total, specifically looked at benzene exposure and are described in Table 1. Of these 22 studies, 17 (77%) show some elevation of NHL risk, with 8 of 22 (36%) finding statistically significant associations between NHL risk and benzene exposure. The eight studies showing increased risks are diverse in nature and include the recent study of Miligi, Vineis, Costantini, and colleagues, who described an increased risk of NHL with exposure to aromatic hydrocarbon solvents of 2.1 (1.1-4.3; ref. 5). Medium/high exposure to benzene-containing solvents elevated the risk of NHL [odds ratio (OR), 1.6; 95% confidence interval (95% CI), 1.0-2.4], especially diffuse NHL (OR, 2.4; 95% CI, 1.3-4.5; Table 1; ref. 5). They update and extend their findings elsewhere in this issue.

Dryver et al. (6) found increased risks of NHL for exposures to aromatic hydrocarbon solvents and gasoline (OR, 1.72; 95% CI, 1.10-2.71; ref. 6), including dose-dependent effects for exposure to aromatic hydrocarbon solvents and duration-dependent effects for gasoline. Risks were also significantly increased for employment as an automobile mechanic, gas station attendant, and painter or varnisher. Blair et al. (7) observed that increased risks of diffuse and follicular NHL were associated with increased benzene exposure intensity and were significantly increased among both painters and persons with metal exposures.

Fabbro-Perray et al. (8) identified an association between self-reported benzene exposure and increased risk of NHL (OR 2.0; 95% CI, 1.1-3.9) that was especially strong for those exposed for >810 days (Table 1). Hardell et al. (9) reported increased risks of lymphoma in workers with high exposures to benzene, styrene, and chlorinated solvents (RR, 4.5; 95% CI, 1.9-11.4). Following up this study, Hardell et al. (10) observed large increases in NHL risk (OR, 28; 95% CI, 1.8-730) subsequent to benzene exposure; smaller, but significant, risks of NHL were observed for exposure to degreasers and organic solvents (Table 1).

Ott et al. (11) found elevated risks of NHL in foremen and others who had long careers in maintenance and construction work at chemical facilities. In particular, male workers with >5 years of occupational exposure to benzene had a 1.6-fold elevated risk of NHL. Xu et al. (12) found statistically significant excesses of lymphoma for exposure to benzene (adjusted OR, 2.78; P = 0.001) in a hospital-based study in China.

Petroleum refinery workers are potentially at risk of lymphoma and other cancers from exposure to benzene concentrated during the refining process. There have been many studies of workers in this industry, most of which have been conducted by industry-funded research groups. We identified 26 studies of petroleum refinery workers reporting morbidity or mortality for lymphomas and all (combined) neoplasms (Table 2). Lymphoma mortality or morbidity is significantly elevated in a few studies, but in most studies the rates are slightly but not significantly elevated. These slightly elevated rates, however, are striking because the studies report the population to have a deficit of mortality or morbidity from all (combined) neoplasms or all causes. Deficits in mortality and morbidity in cohorts of workers are often identified as the “healthy-worker effect.” The healthy-worker effect, however, would predict that the workers experience deficits in mortality and morbidity from all causes: morbidity and mortality rates near those experienced by the general population or white-collar workers are suggestive of adverse health effects from occupational exposures. Among the 26 studies reporting mortality or morbidity from all neoplasms and from NHL, 23 (88%) showed that the rate of lymphoma morbidity or mortality was higher than that for all neoplasms. A similar trend appears for studies reporting all-cause mortality: the rate of lymphoma mortality is greater than the rate of all-cause mortality in 14 of 17 (82%) of studies reporting both outcomes.

We explored this remarkably consistent finding further by adjusting for the healthy-worker effect in the study of Dagg et al. (13). Miettinen and Wang (14) proposed analyzing proportionate mortality studies as case-control studies to yield an odds ratio estimate. The same approach can be used to adjust for the healthy-worker effect. The standardized mortality ratio for lymphatic and hematopoietic cancers was not significantly elevated at 107 (95% CI, 86-132) without adjustment in this study, but the standardized mortality ratio for all causes of death was 73 (95% CI, 71-76), showing clear evidence of a healthy-worker effect. To adjust for the health worker effect, we considered lymphatic and hematopoietic cancers to be cases, and all other causes of death excluding lymphatic and hematopoietic cancers as controls. The OR for lymphatic and hematopoietic cancers estimated using Miettinen's method (14) was 1.48 (95% CI, 1.20-1.83; P < 0.001), providing clear evidence that the healthy worker effect masked increased blood cancer risks. Many cohort studies of refinery workers in Table 2, like the study by Dagg et al. (13), show reduced standardized mortality ratios for all neoplasms combined, suggesting that masking of NHL risk by the healthy-worker effect is common. A comprehensive reevaluation of these studies with appropriate adjustments should be undertaken.

A study of benzene-exposed workers in China, conducted by investigators from the U.S. National Cancer Institute and the Chinese Academy of Preventive Medicine, is ongoing (15). To date, the primary NHL-related finding from the ∼74,000 observed workers is a 3-fold increase in NHL risk for benzene-exposed workers, increasing to a 4-fold excess in NHL risk for workers exposed to benzene for 10 or more years. For workers in the chemical industry in this cohort, the relative risk (RR) was even higher at 7.8. In a study of chemical industry workers in the United States, Wong (16) found that the RR for NHL mortality among white benzene-exposed workers was 8.6 (P = 0.02) compared with unexposed workers. For workers continuously exposed to benzene, the RR was 9.6 compared with unexposed workers. Cohort studies of workers in the chemical manufacturing industry are summarized in Supplementary Table S2.

Studies of worker cohorts in the rubber industry, like the petroleum refining and chemical industries, have found elevated risks of lymphatic and hematopoietic malignancies (Supplementary Table S3). The role of benzene in these cancers, however, can be difficult to define because workers in this industry are often exposed to other potential carcinogens, such as 1,3-butadiene. The only study of rubber workers in which it is widely agreed that the only significant solvent exposure was benzene is the Pliofilm study of rubber hydrochloride workers (17, 18). Although significantly elevated risks for leukemia and multiple myeloma were observed, an elevated risk for NHL was not. The failure of this study to identify an association between benzene exposure and NHL mortality, however, does not exclude the possibility of the association. In particular, the low mortality and long latency of NHL severely limited the power of the Pliofilm study to detect an association between benzene-exposure and NHL mortality. We note that the 5-year survival of NHL and lymphocytic leukemias exceeds 60%, compared with <20% for acute leukemia and myeloma. This explains why the expected number of deaths from NHL in the Pliofilm cohort was less than those expected from acute leukemia, although the incidence of NHL is four to five times that of leukemia. Further, the typical latency period for NHL is likely to be much longer than that for acute leukemia. Thus, the Pliofilm study has limited power to detect elevated risks of NHL following benzene exposure.

B-cell chronic lymphocytic leukemia is now classified together with small lymphocytic lymphoma as a form of NHL. A number of studies show elevated risks of chronic lymphocytic leukemia at relatively low levels of benzene exposure (Supplementary Table S1). The most notable is that of Glass et al. (19), but others have also showed elevated risks including studies in the United Kingdom of petroleum distribution workers (20).

Numerous studies have reported associations between benzene exposure and the induction of lymphomas in mice. Studies by Cronkite et al. (21) and Snyder et al. (22) in the early 1980s showed inhalation of benzene caused excess lymphomas in various mouse strains. The 1986 National Toxicology Program carcinogenicity bioassay of benzene also reported an excess of malignant lymphomas, as well as other cancers, in male and female B6C3F1 mice. Studies in Italy by Maltoni et al. (23) showed excesses of lymphosarcomas in RF/J mice. As with many rodent toxicology studies, there has been debate over the relevance of these findings to humans.

There are many similarities between leukemia induced by benzene and that induced by chemotherapy with alkylating agents, so-called therapy-related leukemia, and the latter has been proposed as a model of chemical-induced leukemogenesis (24). Treatment of primary cancers with alkylating agents such as melphalan dramatically increases risks of secondary leukemias. Interestingly, as discussed by Krishnan and Morgan in this issue, numerous studies show that alkylating agent chemotherapy also increases the risk of secondary NHL as well as leukemia. Thus, benzene may act like alkylating agents and ionizing radiation in inducing both leukemias and lymphomas.

A comprehensive review of all case-control and cohort studies that identified probable occupational exposures to benzene and NHL morbidity or mortality showed evidence of an association between occupational benzene exposure and NHL. Studies in experimental animals and of therapy-related lymphoma in the clinical setting support this finding. Potential mechanisms for lymphoma induction by benzene include immunotoxicity and chromosomal damage resulting in translocations and deletions. Further work is needed to elucidate these mechanisms.