Smoking and diabetes are the only established risk factors for pancreatic cancer. Findings from recent studies suggest that obesity may also be associated with an increased risk of pancreatic cancer, but several earlier studies were less conclusive. We examined this relationship in a meta-analysis of published data. Six case -control and eight cohort studies involving 6391 cases of pancreatic cancer were identified from a computer-based literature search from 1966 to 2003. The relative risk per unit increase in body mass index was estimated for each of the studies from the published data. In a random effects model, the summary relative risk per unit increase in body mass index was 1.02 (95% CI: 1.01 -1.03). There was some evidence of heterogeneity between the studies' results (P ¼ 0.1). The summary relative risk estimates were slightly higher for studies that had adjusted for smoking and for case -control studies that had not used proxy respondents. The estimated per unit increase in body mass index would translate into a relative risk of 1.19 (95% CI: 1.10 -1.29) for obese people (30 kg m À2 ) compared to people with a normal body weight (22 kg m À2 ). These results provide evidence that the risk of pancreatic cancer may be weakly associated with obesity. However, the small magnitude of the summary risk means the possibility of confounding cannot be excluded. British Journal of Cancer (2003) Pancreatic cancer is the fifth most common cause of cancer mortality in the European Union and North America and is responsible for approximately 70 000 deaths in these regions annually (Ferlay et al, 2001). Smoking and diabetes are the only well-established risk factors for this invariably fatal disease (Gapstur and Gann, 2001). Smoking, however, is thought to account for at most 20 -40% of pancreatic cancer cases (International Agency for Research on Cancer, 1990). The findings from several large US cohort studies suggest that obesity (defined as body mass index of at least 30 kg m À2 ) may also be a risk factor for pancreatic cancer (Michaud et al, 2001;Calle et al, 2003). The results of several earlier observational studies were less conclusive. The aim of this meta-analysis was to examine the relationship between pancreatic cancer and body mass index in relevant published epidemiological studies.
MATERIALS AND METHODS
Data sourcesEpidemiological studies were identified through searches of the electronic databases MEDLINE (1966MEDLINE ( -2003, EMBASE (1980EMBASE ( -2003 and the Science Citation Index (1981 -2003) and also from citations in the selected papers and review articles. The key words that were used for the search were pancreatic cancer, obesity, body mass index, anthropometric factors and risk factors. The search was limited to human studies and restricted to peer-reviewed articles. No language or date limitations were imposed.
Study selectionEach study was required to have published information on the number of study participants and on age-adjusted or age-matched relative risks (or odds ratios, subsequen...