Chin-Hsiao Tseng1. 1. Department of Internal Medicine, College of Medicine, National Taiwan University, Taipei, Taiwan. ccktsh@ms6.hinet.net
Abstract
OBJECTIVE: This study aimed to investigate whether the reported relationship between diabetes and pancreatic cancer (PC) could result from detection bias and whether dyslipidemia and/or new-onset diabetes (diagnosed within 1 year) could predict PC. METHODS: A random sample of 1 million subjects covered by National Health Insurance was recruited. From 2003 to 2005, 495,493 men and 503,901 women without PC were followed up. Cox regression was used to evaluate the adjusted relative risk considering potential PC detection examinations and covariates. RESULTS: Diabetic patients had a significantly higher probability of receiving examinations that might lead to PC diagnosis. In Cox proportional hazards regression models, diabetes was not a significant predictor, but dyslipidemia was significantly associated with an approximately 40% higher risk of PC. Age, living in more urbanized regions, and potential PC detection examinations were significant covariates. Patients with new-onset diabetes and previous dyslipidemia had a remarkably higher risk compared with those without either condition (relative risk [95% confidence interval], 2.512 [1.169-5.398]). CONCLUSIONS: Dyslipidemia, but not diabetes, is a significant risk factor for PC. The link between diabetes and PC is likely due to confounders and detection bias. Patients with new-onset diabetes and a history of dyslipidemia are at an especially high risk of PC.
OBJECTIVE: This study aimed to investigate whether the reported relationship between diabetes and pancreatic cancer (PC) could result from detection bias and whether dyslipidemia and/or new-onset diabetes (diagnosed within 1 year) could predict PC. METHODS: A random sample of 1 million subjects covered by National Health Insurance was recruited. From 2003 to 2005, 495,493 men and 503,901 women without PC were followed up. Cox regression was used to evaluate the adjusted relative risk considering potential PC detection examinations and covariates. RESULTS:Diabeticpatients had a significantly higher probability of receiving examinations that might lead to PC diagnosis. In Cox proportional hazards regression models, diabetes was not a significant predictor, but dyslipidemia was significantly associated with an approximately 40% higher risk of PC. Age, living in more urbanized regions, and potential PC detection examinations were significant covariates. Patients with new-onset diabetes and previous dyslipidemia had a remarkably higher risk compared with those without either condition (relative risk [95% confidence interval], 2.512 [1.169-5.398]). CONCLUSIONS:Dyslipidemia, but not diabetes, is a significant risk factor for PC. The link between diabetes and PC is likely due to confounders and detection bias. Patients with new-onset diabetes and a history of dyslipidemia are at an especially high risk of PC.