The present work reviews the available epidemiological data on the relationship of pancreatic cancer and smokeless tobacco used in North America and Europe. Meta-analytic methods were used to estimate the overall relative risk. Four cohort and seven case-control studies were identified. Two sets of estimates were selected for meta-analysis. The first meta-analysis was based on study-specific estimates for the overall population of smokers and nonsmokers combined, where they were available, otherwise on estimates for never smokers. The overall random-effects risk estimate was found to be 1.10 (95% confidence interval 0.73-1.65). Sub-group analyses by continent, study type, or type of smokeless tobacco also showed no increased risk. The second meta-analysis was based on study-specific estimates for never smokers, where they were available, otherwise on combined smokers/nonsmokers estimates. The overall random-effects risk estimate was found to be 1.15 (95% confidence interval 0.67-1.98). There were, however, some indications of increased risk in cohort studies, in studies conducted in Sweden or Norway, and in studies of snuff. As there are a number of weaknesses in the currently available studies, no definite conclusion can be drawn. Our results are less indicative of a causal relationship than other reviewers have suggested; additional studies will have to be conducted to clarify the situation.