Diabetes has been postulated to be both a risk factor and a consequence of pancreatic cancer, but the degree of risk and associated clinical factors remain unclear.

We conducted a population-based case-control study of pancreatic cancer in the San Francisco Bay Area between 1995 and 1999. Rapid case ascertainment through the Surveillance, Epidemiology and End Results registry for cases and random selection from the general population for controls were employed to identify study participants with no proxy interviews.

Five hundred thirty-two cases and 1,701 controls were interviewed. Participants with pancreatic cancer were more likely to report a history of diabetes (13%) than were controls [9%; odds ratio (OR), 1.5; 95% confidence interval (95% CI), 1.1-2.1]. Compared with diabetics in the control group, diabetics in the case group had a shorter duration of diabetes (P = 0.0003) and a larger proportion of insulin users (P = 0.002). Risk for pancreatic cancer varied with duration of diabetes (OR, 2.4; 95% CI, 1.4-4.0 for 1-4 years; OR, 2.0; 95% CI, 1.2-3.4 for 5-9 years; and OR, 0.86; 95% CI, 0.52-1.4 for >or=10 years diabetes duration; P(trend) = 0.004). Among diabetics, use of oral diabetes medication or insulin for >or=5 years was not associated with pancreatic cancer, but insulin use of <5 years was associated with a 6.8-fold risk for pancreatic cancer (95% CI, 3.7-12).

Recent-onset diabetes may be a complication or an early marker of pancreatic cancer. Diabetes of short duration with insulin use conferred a substantially elevated risk for pancreatic cancer and may reflect insulin resistance that is elicited by pancreatic cancer.

Although many individuals with pancreatic cancer have diabetes, the association between new-onset diabetes mellitus and the subsequent incidence of pancreatic cancer is unclear.

We conducted a retrospective cohort study to estimate the incidence of pancreatic cancer subsequent to a new diabetes diagnosis and to evaluate factors associated with a subsequent pancreatic cancer diagnosis. We used the Veterans Health Administration National Patient Care Database to assemble a cohort of 1,421,794 US veterans without prior diabetes or pancreatic cancer diagnoses. We recorded coding for new diabetes diagnoses (> or =2 International Classification of Diseases-9 codes for diabetes within a 12-month period), pancreatic cancer, age, sex, race, and common gastrointestinal symptoms.

A total of 36,631 (2.6%) of the 1,421,794 veterans were diagnosed with new-onset diabetes in 1999; 149 subsequently received a diagnosis of pancreatic cancer. Pancreatic cancer incidence in patients with new-onset diabetes (83.8/100,000 person-years) was 2.2-fold higher (95% confidence interval, 1.84-2.56) than in nondiabetics, and was highest during the first 2 years after diabetes diagnosis. One additional pancreatic cancer was diagnosed for every 332 new diabetics over 6 years. A subsequent pancreatic cancer diagnosis (among new-onset diabetics) was associated independently with younger age groups, changes in bowel habits, constipation, epigastric pain, and malnutrition.

New-onset diabetes was associated with a significantly increased rate of pancreatic cancer diagnosis, particularly in the first 2 years after diabetes diagnosis. Factors associated with pancreatic cancer diagnosis included younger age groups and the presence of gastrointestinal symptoms. The absolute incidence of pancreatic cancer was low.

Despite accurate diagnosis, better radiologic techniques, and safer surgery, long-term survival after surgical therapy for pancreatic cancer is disappointing. Median survival following pancreaticoduodenal resection is 12 to 15 months independent of surgical expertise, hospital size, or technical factors. Subsets of favorable tumors and longer survival times after surgery have been defined and include: small tumor size and low-grade lesions, tumor-free margins, and absence of nodal, venous, or perineural invasion; however, long-term survivors of pancreatic cancer may have none of these favorable features, and their tumors commonly manifest the most adverse tumor prognostic features. The converse that small-sized, histologically favorable tumors result in long-term survivors, also is not true. Five-year survival rates average 5% or less after all resections. In a large series in which 118 pancreatic resections were performed in 684 evaluated patients over a 6-year period, there were 12 5-year survivors, 5 of whom died in the sixth year. A report of 10-year survivors after surgery numbered 13 patients. The best actual 5-year survival rate was reported by Trede et al. Of the 37, 5-year survivors from a cohort of 118 patients, more than half died of cancer. This far exceeds any other actual survival rate and may be explained by a smaller tumor size. Farnell et al reported a 5-year survival rate difference (i.e., actuarial survival) in a subset of 174 resected patients with adenocarcinoma without perineural or duodenal invasion and with negative nodes (23% versus 6.8%), respectively. An impressive, large series of 616 patients with resected adenocarcinoma of the pancreas who underwent PDR (85%), distal pancreatectomy (9%), and total pancreatectomy (6%), has been reported. The mortality rate was 2.1%, and postoperative complications occurred in 30%. The five-year survival rate was 15%. The author's best result was observed among 20 initially "unresectable" patients who were treated with chemoradiation therapy, followed by tumor extirpation. Among the 18 surgical survivors there are seven five-year survivors, three of whom are in their tenth year of survival. They are discussed in the article by Cooperman et al ("Long-term Follow-up...") elsewhere in this issue.

Patients with pancreatic adenocarcinoma have a high incidence of diabetes, profound insulin resistance, and high circulating amylin concentrations. It was hypothesized that pancreatic cancer cells produce a factor that stimulates islets to secrete amylin but not insulin.

Amylin and insulin secretion were measured after coculture of pancreatic cancer cells with beta cells (BRIN-BD11). The factor responsible was characterized by exposing beta cells to cancer cell-conditioned medium.

Coculture with pancreatic (PANC-1 HPAF, and MiaPaCa2) but not colonic cancer cells (Colo 320) significantly increased amylin secretion but did not change insulin output. This effect was both time and cell number dependent. Coculture with PANC-1 or HPAF cells significantly decreased intracellular amylin, but not insulin, content. PANC-1 or HPAF cell-conditioned medium also increased amylin secretion and decreased intracellular amylin content. The factor responsible was extracted under both neutral and acidic conditions, was heat labile, and had a molecular weight of approximately 1500.

A soluble factor from pancreatic cancer cells selectively stimulates amylin secretion from islet cells, explaining the excessive amylin secretion found in pancreatic cancer. Because elevation of amylin concentration is an early feature of pancreatic cancer, characterization and measurement of the tumor-derived amylin-releasing factor might be valuable in the early detection of this disease.

This article reviews the epidemiology of cancer of the pancreas, both descriptive and analytical, at all times cognizant of the problems of misdiagnosis, particularly underdiagnosis, of this lethal disease that continue to hinder epidemiological studies. Pancreas cancer is consistently reported to occur more frequently in men than in women, in blacks than in whites, and in urban rather than rural population groups. In some countries, the mortality rates continue to rise, whereas in others, declining levels of disease can be seen among members of younger birth cohorts. Although some of these patterns can be explained by variation in pancreas cancer risk factors, many cannot. Analytical studies consistently demonstrate that cigarette smoking increases the risk of cancer of the pancreas, and this appears, at the present time, to be the only clearly demonstrated risk factor for pancreatic cancer. Although the association with disease risk and coffee consumption, alcohol consumption, occupational exposures, diabetes, pancreatitis, and other factors requires clarification, it appears likely that the most fruitful research area in the coming years may involve exploration of pancreatic cancer risk and nutritional practices.