Petroleum workers are exposed to benzene or benzene-containing petroleum products. As such, studies of these workers provide an opportunity for investigating the relationship between benzene and non-Hodgkin's lymphoma (NHL). However, few cohort studies of petroleum workers report results of NHL separately. One reason is that NHL is usually grouped with other lymphopoietic cancers in the analysis. Another reason is the relatively small number of NHL cases in some studies. To determine the risk of NHL in petroleum workers, we identified 26 cohorts of petroleum workers in the United States, the United Kingdom, Canada, Australia, Italy, and Finland. Authors of the original studies were contacted, and data on the number of observed deaths and person-years of observation were requested. Data from these studies were reviewed individually as well as combined in a pooled analysis (meta-analysis). In particular, results for individual cohorts, most of which had never been reported before, were presented. The combined multinational cohort consisted of more than 308,000 petroleum workers (6.6 million person-years), and the observation period covered an interval of 60 years from 1937 to 1996. A total of 506 NHL deaths were observed, compared with 561.68 expected. The standardized mortality ratio was 0.90 and the 95% confidence interval was 0.82 to 0.98. Analyses were performed by type of facility and industrial process. Stratum-specific standardized mortality ratios (95% confidence intervals) were 0.96 (0.86 to 1.07) for US refinery workers, 1.12 (0.90 to 1.37) for non-US refinery workers, 0.64 (0.50 to 0.82) for product (gasoline) distribution workers, and 0.68 (0.47 to 0.95) for crude oil workers. When individual cohorts were stratified by length of observation, no pattern was detected. In general, exposure levels before 1950 were much higher than thereafter. However, analysis of workers by hire date (< 1950, > or = 1950) revealed no difference in NHL mortality. Furthermore, none of the individual studies showed significant exposure-response relations. In summary, results from individual studies, as well as from the pooled analysis, indicated that petroleum workers were not at an increased risk of NHL as a result of their exposure to benzene or other benzene-containing petroleum products in their work environment. This conclusion was supported by cohort studies of workers in other industries who were exposed to benzene as well as by population-based case-control studies of NHL and occupational exposures.