Diabetes mellitus and pancreatic cancer mortality in a prospective cohort of United States adults.

Diabetes mellitus and pancreatic cancer are known to be associated, but it is not known whether diabetes is a true risk factor, preceding development of the cancer, or if it is an early manifestation of the cancer. To address this uncertainty, we examined the association of pancreatic cancer mortality and reported diabetes of at least one year's duration in a large, prospective study of United States adults. The vast majority of diabetes in this cohort is likely to be non-insulin-dependent diabetes. After 12 years of follow-up, 2,953 deaths from pancreatic cancer were observed in a cohort of 1,089,586 men and women who were cancer-free at study entry in 1982. Cox proportional hazards models, adjusted for age, race, smoking, family history of pancreatic cancer, body mass index (wt/ht2), and education, were used to assess associations. A history of diabetes was significantly related to pancreatic cancer mortality in both men (rate ratio [RR] = 1.49, 95 percent confidence interval [CI] = 1.25-1.77) and women (RR = 1.51, CI = 1.24-1.85). However, the strength of the association varied over the follow-up period. The death rate from pancreatic cancer was twice as high in diabetics as in non-diabetics during the second and third years of follow-up (adjusted RR = 2.05, CI = 1.56-2.69) but only about 40 percent higher in years nine to 12 (adjusted RR = 1.38, CI = 1.08-1.77). The small but persistent increased risk of death from pancreatic cancer, seen even when the diagnosis of diabetes preceded death by many years, supports the hypothesis that diabetes may be a true, albeit modest, risk factor for pancreatic cancer.