Diabetes, insulin use, smoking, and pancreatic cancer mortality in Taiwan.

The aim of the study was to evaluate the link between diabetes and pancreatic cancer (PC) mortality and the joint effect of smoking and insulin use on PC mortality. A total of 39,988 men and 46,909 women with type 2 diabetes, aged ≥25 years and recruited in 1995-1998, were followed to 2006 for PC mortality. Age-sex-specific mortality rate ratios for diabetic patients versus the general population were calculated. Cox regression was used to evaluate hazard ratios for PC mortality for covariates including age, sex, diabetes duration, body mass index, smoking, insulin use, and area of residence. The interaction and joint effect of smoking and insulin use were also evaluated. A total of 89 men and 63 women died of PC. The mortality rate ratios (95 % CI) showed a significantly higher risk in diabetic patients with a magnitude most remarkable at the youngest age: 1.51 (1.15, 1.98), 2.02 (1.35, 3.03), and 8.36 (5.39, 12.98) for ≥65, 55-64, and 25-54 years old, respectively, for men; and 1.16 (0.84, 1.59), 2.12 (1.39, 3.23) and 3.33 (1.14, 9.68), respectively, for women. In multivariable Cox regression analysis, only age was significantly predictive for PC mortality. Although smoking and insulin use might be associated with a 50 % higher risk when analyzed as individual risk factors, they did not reach statistical significance. The interaction term of smoking and insulin use was also not statistically significant in additional modeling. However, smoking and insulin use jointly increased the risk with an adjusted hazard ratio (95 % CI) of 3.04 (1.37, 6.73) when compared to patients who did not smoke and did not use insulin. Diabetic patients have a significantly higher risk of PC mortality. In patients with type 2 diabetes, smoking and insulin use may jointly increase the risk by threefold.