Epidemiological studies clearly indicate that the risk of pancreatic cancer (PC) is increased in diabetic patients, but most studies focus on overall diabetes or type 2 diabetes mellitus (T2DM), and there are few studies on the risks of type 1 and type 3c (secondary) diabetes. Possible mechanisms for increased cancer risk in diabetes include cellular proliferative effects of hyperglycemia, hyperinsulinemia, and abnormalities in insulin/IGF receptor pathways. Recently, insulin and insulin secretagogues have been observed to increase the PC risk, while metformin treatment reduces the cancer risk in diabetic subjects. In addition, anticancer drugs used to treat PC may either cause diabetes or worsen coexisting diabetes. T3cDM has emerged as a major subset of diabetes and may have the highest risk of pancreatic carcinoma especially in patients with chronic pancreatitis. T3cDM is also a consequence of PC in at least 30% of patients. Distinguishing T3cDM from the more prevalent T2DM among new-onset diabetic patients can be aided by an assessment of clinical features and confirmed by finding a deficiency in postprandial pancreatic polypeptide release. In conclusion, diabetes and PC have a complex relationship that requires more clinical attention. The risk of developing PC can be reduced by aggressive prevention and treatment of T2DM and obesity and the prompt diagnosis of T3cDM may allow detection of a tumor at a potentially curable stage.
Pancreatic cancer (PC) is one of the most lethal malignant diseases due to the high rate of advanced stage disease at diagnosis and the lack of any effective medical therapy (Ko & Tempero 2009). The overall incidence of the disease has increased over the past few decades such that over 265 000 people worldwide are diagnosed annually (Jemal et al. 2011). Five-year survival after surgical resection has slowly risen to about 20%, owing to better selection of surgical candidates and some modest improvements in adjuvant therapy, but only 10–15% of patients are candidates for resection, and more than 95% of all patients will succumb to their illness within 2 years of diagnosis (Edwards et al. 2005). This high mortality rate confers PC as the fourth leading cause of cancer-related deaths in the USA currently (Jemal et al. 2010). With the continuing decline in mortality rates of cancers of the lung, colon, breast, and prostate due to the combined effects of widespread screening, smoking cessation, and more effective therapy, it is projected that pancreas cancer will become the leading cause of cancer-related deaths in the USA by 2050 (Siegel et al. 2012).
The etiology of PC is complex and poorly understood. Therefore, the identification of risk factors, especially those that are modifiable through medication or behavioral change, is important for preventing the development and progression of PC. Risk factors for PC include family history, smoking, obesity, chronic pancreatitis (CP), and diabetes mellitus (DM). The recent increase in the prevalence of type 2 DM (T2DM) is thought to have contributed to a parallel rise in the incidence of PC. Roughly half of all PC patients are found to have DM at the time of diagnosis, and roughly half of the DM that is present at the time of PC diagnosis is of new onset having developed over 2–3 years preceding the diagnosis of PC. This new-onset DM is therefore thought to be secondary, or T3cDM. The association between PC and DM has been investigated extensively, but the causal relationships have yet to be fully elucidated, in part due to difficulties in distinguishing T2DM from T3cDM.
In this review, we highlight the important recent studies and discuss the available evidence concerning the possible mechanisms that are involved in the etiological role of DM in the development of PC, the effects of antidiabetic therapy on the risk of PC, and a strategy for distinguishing T3cDM from T2DM so as to identify the patient with early and indolent PC.
Diabetes (and obesity) as a cause of PC
The risk of PC is increased in T1DM, T2DM, and T3cDM
It has long been known that PC is associated with DM, and recent studies have revealed that about 85% of patients diagnosed with PC have impaired glucose tolerance or frank DM (Pannala et al. 2008; Fig. 1). DM is a group of metabolic disorders characterized by hyperglycemia. The three most common subtypes of DM differ greatly in their metabolic and hormonal characteristics, however (Table 1; Cui & Andersen 2011). T1DM is associated with a profound or absolute deficiency of endogenous insulin secretion and an absolute requirement for exogenous insulin administration. Hyperglycemia and hyperinsulinemia coexist in T2DM due to insulin resistance in peripheral tissues usually in association with obesity. T3cDM is associated with benign and malignant disease of the exocrine pancreas, including acute and CP of any etiology, pancreatic neoplasms, pancreatic trauma, pancreatic resection, cystic fibrosis, hemochromatosis, fibrocalculous pancreatopathy, and pancreatic agenesis and is characterized by a severe deficiency of all pancreatic glucoregulatory hormones (American Diabetes Association 2011, Cui & Andersen 2011).
Clinical and laboratory findings in types of diabetes mellitus
|Type 1||Type 2||Type 3c|
|Peripheral insulin sensitivity||Normal or increased||Decreased||Increased|
|Hepatic insulin sensitivity||Normal||Normal or decreased||Decreased|
|Glucagon levels||Normal or high||Normal or high||Low|
|PP levels||Normal or low (late)||High||Low|
|GIP levels||Normal or low||Normal||Low|
|GLP1 levels||Normal||Normal or low||Normal or high|
|Typical age of onset||Childhood or adolescence||Adulthood||Any|
IDDM, insulin-dependent diabetes mellitus; NIDDM, non-insulin-dependent diabetes mellitus; PP, pancreatic polypeptide; GIP, glucose-dependent insulinotropic polypeptide; GLP1, glucagon-like peptide 1. Modified, with permission, from Slezak LA & Andersen DK 2001 Pancreatic resection: effects on glucose metabolism. World Journal of Surgery 25 452–460. Copyright 2001 International Society of Surgery.
Epidemiological studies have demonstrated that DM is a risk factor for multiple forms of malignancy including PC (Wideroff et al. 1997, Calle et al. 2003, Huxley et al. 2005). A recent meta-analysis of 35 cohort studies concluded that a twofold risk of pancreatic malignancy exists in diabetic patients (Ben et al. 2011). DM was associated with an increased relative risk (RR) of PC (RR=1.94; 95% confidence interval (95% CI): 1.66–2.27), with significant evidence of heterogeneity among the studies surveyed (P<0.001, I2=93.6%). Subgroup analyses revealed that the increased risk of PC was independent of geographic location, sex, study design, alcohol consumption, body mass index (BMI), and/or smoking status. In addition, the risk of PC correlated negatively with the duration of DM, with the highest risk of PC found among patients with DM diagnosed within <1 year. This finding implies that many of the diabetic subjects had PC-induced DM (T3cDM), but the type of DM was not identified in most epidemiological studies.
A meta-analysis of three recent studies in which 2192 PC patients were compared with 5113 controls revealed a 1.8-fold increase in risk of PC associated with T2DM, although many of the patients classified as T2DM likely harbored T3cDM, which became evident close to the time of diagnosis of the cancer (Li et al. 2011). Finally, a recent cohort study comparing 110 919 DM subjects and 211 695 controls provides strong support for an etiological role of T2DM and hyperinsulinemia in the pathogenesis of PC (Yacoub et al. 2011). The DM cohort had 124 PCs developed during 0.558 million person-years (MPY) of follow-up (222 PCs/MPY), whereas the control cohort had 140 PCs developed during 1.299 MPY of follow-up (108 PCs/MPY). Stratified Cox regression of PC incidence yielded a hazard ratio (HR) of 2.17 (95% CI: 1.70–2.77) for T2DM compared to controls. So the risk of developing PC is clearly although modestly increased (about twofold) in the presence of long-standing, mostly T2DM.
Genetic studies have also provided insights underlying the association of DM and PC. Pierce et al. (2011) examined the 37 risk alleles of T2DM and found two that showed nominally significant positive associations with PC risk (FTO rs8050136 per allele odds ratio (OR)=1.12; 95% CI: 1.02–1.23; MTNR1B rs1387153 OR=1.11; 95% CI: 1.00–1.23), and the glucose-raising allele of MADD rs11039149 was associated with increased risk of PC (OR=1.14; 95% CI: 1.03–1.27). Prizment et al. reported that glucokinase regulator (GCKR) rs780094, a single-nucleotide polymorphism related to T2DM, may be associated with PC risk. In a multivariate-adjusted model, a significant association was observed only for rs780094 in the GCKR gene: ORs for PC were 1.00 for the TT genotype, 1.35 (95% CI: 0.71–2.58) for the CT genotype, and 2.14 (95% CI: 1.12–4.08) for the CC genotype (P=0.01), which did not change after the adjustment for the presence of DM (Prizment et al. 2012).
There are few studies that specifically address cancer incidence in T1DM patients. A cohort study evaluating cancer incidence in nearly 30 000 Swedish T1DM patients diagnosed in the period 1965–1999 identified 355 cases of cancer, which equated to a standardized incidence ratio of 1.2 (95% CI: 1.0–1.30) compared with the general Swedish population (Zendehdel et al. 2003). In contrast to these modest findings, a meta-analysis including three cohort studies and six case–control studies has found that the RR for PC was doubled in T1DM patients and ‘young-onset’ diabetics in comparison with nondiabetics (Stevens et al. 2007).
T3cDM has emerged as a major subset of the total population of DM and harbors the highest risk of PC especially in patients with T3cDM secondary to CP (Cui & Andersen 2011). Persons with any form of CP are at increased risk of developing PC, and Lowenfels et al. (1993) observed that the risk is cumulative over the course of CP, such that 4–5% of patients develop PC over the course of 20 years, a risk that is 10–20 times greater than the general population. In a case–control study in which 823 PC patients and 1679 controls were surveyed from centers in Australia, Canada, The Netherlands, and Poland, the risk of PC was seen to be increased (RR 4.7) for subjects with a history of pancreatitis (not otherwise defined), although was increased only among alcohol drinkers (RR 7.5) and not among teetotalers (RR 0.57; Maisonneuve et al. 2010; Fig. 2). A recent cohort study in Taiwan showed that in addition to age, CP (HR=19.40), gallstones (HR=2.56), and hepatitis C infection (HR=3.08) were significant factors predicting PC; patients with concurrent DM and CP had a dramatically elevated risk of developing PC (HR=33.52), compared with subjects without these comorbidities (Liao et al. 2012).
Obesity as a risk factor for PC
In 2003, Calle et al. (2003) examined the American Cancer Society database to assess the possible role of obesity on the risk of cancer. In their cohort of over 900 000 individuals, massive obesity (BMI >40) was associated with a 50–60% increased death rate from cancers of the pancreas, liver, kidney, colon and rectum, esophagus, non-Hodgkins lymphoma, and multiple myeloma over a 16-year period. The greatest effects were seen in liver cancer (RR 4.52 for men and 1.68 for women), colorectal cancer (RR 1.84 for men and 1.46 for women), and PC (RR 1.49 for men and 2.76 for women). Over 80% of T2DM patients are obese and studies on the obesity–PC association are influenced by the high prevalence of obesity in T2DM patients and by the high percentage of T2DM that is undiagnosed. Recent epidemiological studies have demonstrated that a high BMI is positively associated with an increased risk of many common cancers independent of the coexistence of T2DM (Teucher et al. 2010), and a high BMI has been identified as an independent risk factor for PC (Michaud et al. 2001, Silverman 2001, Gumbs 2008, Arslan et al. 2010). More recently, large prospective cohort studies with a long duration of follow-up have been conducted in the USA showing a positive association between high BMI and the risk of PC (adjusted RR 1.13–1.54; Jiao et al. 2010, Johansen et al. 2010), suggesting the role of obesity and overweight as conferring a higher risk for the development and eventual death due to PC. Among the many possible mechanisms involved, hyperinsulinemia, diet and nutritional factors, and other hormone abnormalities have been suggested as causal factors.
Reduced caloric intake and physical exercise have both been shown to reduce the risk of PC (Michaud et al. 2001, Silverman 2001), and studies on the effects of bariatric surgery are also consistent with a reduction in PC risk after weight loss. In 2007, the Swedish Bariatric Surgery Trial reported a 38% reduction in cancer-related deaths compared with nonoperated controls of similar BMI (Sjöström et al. 2007), and a significant reduction in cancer-related medical care was documented in a study of 1035 bariatric surgery patients compared to 5746 morbidly obese controls reported by Christou et al. (2008). In a similar but larger study, Adams et al. (2009) found a lower incidence of obesity-related cancers, defined as esophageal, colorectal, pancreas, liver, gallbladder, breast, and uterine cancers, in addition to non-Hodgkin's lymphoma, leukemia, and multiple myeloma in 6596 gastric bypass patients compared with 9442 morbidly obese control subjects. As with the studies by Sjöström et al. and Christou et al., the incidence of PC, per se, was too low for valid statistical analysis. Although the mechanisms of risk-reduction remain speculative, the successful treatment of morbid obesity (and concomitant T2DM in about half the subjects) appears to clearly reduce the risk of malignancy and may reduce the incidence of PC.
Proliferation rates of pancreatic ductal endothelium in diabetes and obesity
The effects of DM and obesity on pancreatic ductal pathology were recently studied by Butler et al. (2010), who examined the expression of the neoplastic markers cytokeratin and Ki67 in pancreatic ductal epithelia from 45 human autopsy and nine surgical pathology specimens. In autopsy specimens obtained from obese nondiabetic individuals, pancreatic duct replication was seen to be increased tenfold compared with lean nondiabetics. In lean diabetics, duct epithelial replication was increased fourfold compared with lean nondiabetic subjects. These results indicate the independent effects of obesity and long-standing diabetes on the replication rate in pancreatic ductal cells and presumably therefore on the likelihood of the development of pancreatic exocrine neoplasia (Fig. 3). Markers of pancreatic ductal replication were increased synergistically in obese diabetic subjects. When surgical specimens of CP or nontumor tissue adjacent to PC were examined, even higher rates of the expression of replication markers were seen. These findings support the epidemiological studies that have identified CP, DM, and obesity as contributory to oncogenesis in the pancreas.
Proposed mechanisms of diabetes-related pancreatic carcinogenesis
Although epidemiological studies clearly indicate that DM is positively associated with an increased risk of PC, the molecular mechanism(s) of diabetes-related oncogenesis have not been fully elucidated. Insulin resistance and induced compensatory hyperinsulinemia are widely considered to be likely mechanisms to explain the association of DM and PC (Magruder et al. 2011, Li 2012). Several epidemiological studies have shown that insulin resistance status, characterized by hyperinsulinemia, is associated with an increased risk for a number of malignancies, including carcinomas of the breast, prostate, colon, and kidney. Hyperglycemia has also been shown to be a risk factor for PC. Batty et al. (2004) in England found evidence for a graded dose–response relationship between fasting glucose and the development of pancreatic or liver cancer resulting in mortality. Jee et al. (2005) similarly found a positive linear relationship between fasting glucose and the risk of developing PC across all categories of obesity in a cohort analysis of 1 298 385 Korean patients. In a study of 29 133 Finnish smokers followed for over 10 years, Stolzenberg-Solomon et al. (2005) found that hyperglycemia, hyperinsulinemia, and insulin resistance were each associated with an increased risk of PC. The HR for PC for each entity did not increase significantly until more than 10 years of follow-up, suggesting that long-standing impairments were necessary for the development of PC.
Detailed reviews of these insulin-related mechanisms are provided by Pollak (2008) and Rozengurt et al. (2010). The insulin-like growth factor receptor 1(IGF1R), a tyrosine kinase receptor for IGF1 and 2, has been well documented in cell culture, animal studies, and humans to play a role in malignant transformation, progression, protection from apoptosis, and metastasis in a variety of malignancies. In addition, the hormone insulin and its tyrosine kinase receptor (insulin receptor (IR)) have been documented both in vitro and in vivo to play a key role in cancer biology (Frasca et al. 2008). Chronic hyperinsulinemia is a possible factor favoring cancer initiation and/or progression in diabetic patients due to the mitogenic effect of insulin. Insulin/IGF1 could activate the PI3K/Akt/mammalian target of rapamycin (mTOR) signaling pathway by activation of IR substrates 1–4, which contribute to the development of cancers including PC (Kornmann et al. 1998). Recent reports indicate that IR is overexpressed in several human malignancies and one of the two IR isoforms (IR-A) is especially overexpressed in pancreatic adenocarcinoma. The IRs expressed in malignant tissue also have the capacity to form a hybrid receptor with the IGF1R (Belfiore 2007). By binding to hybrid receptors, insulin may stimulate specific IGF1R signaling pathways, which mediate cell proliferation, inhibition of apoptosis, and growth (Fig. 4). Furthermore, Han et al. (2011) recently found that high glucose promotes PC cell proliferation via the induction of epithelial growth factor (EGF) expression and transactivation of the EGF receptor.
In addition to activation of receptors linked to proliferative and anti-apoptotic pathways, other possible mechanisms have also been suggested as important mediators of pancreatic carcinogenesis. The DM and obesity-related effects on PC development may be mediated by oxidative stress and induced inflammatory responses (Li 2012). Inflammatory responses to external stimuli can accelerate proliferative and repair processes, and obesity itself may activate inflammatory signaling pathways (Greer & Whitcomb 2009, Gallagher & LeRoith 2010). Proinflammatory cytokines released from adipose tissue, known as adipokines, can promote angiogenesis, tumor progression, and metastasis (van Kruijsdijk et al. 2009).
Evidence for increased risk of PC with insulin, insulin secretagogues, and incretin-based treatment
Although the mortality attributable to cancer in DM is overshadowed by that due to cardiovascular disease, emerging data from basic and epidemiological studies suggest that insulin therapy may confer an added risk for cancer, perhaps mediated by signaling through the IGF1R (Azar & Lyons 2010). In 2000, Ding et al. (2000) found that physiological concentrations of insulin increased PC cell proliferation as well as glucose use by activating MAP kinase, PI3 kinase, and GLUT1 expression. In 2003, Bonelli et al. (2003) in Italy reported a case–control study in which the effect of type of treatment for DM on the subsequent development of PC was assessed. Two hundred and forty four patients with documented pancreatic carcinoma and 459 controls were assessed to determine whether insulin therapy or non-insulin therapy affected the subsequent rate of PC development. These investigators found that although DM was associated with a 2.86-fold increase in the risk for PC, the risk increased to 6.49-fold for those treated with insulin, compared with 2.12-fold for those treated with oral hypoglycemic agents. Furthermore, although the duration of insulin treatment had no effect on the high-risk ratio (RR), longer duration of oral hypoglycemic therapy was associated with a lower RR for the development of PC. These findings were corroborated in 2006 by Bowker et al. (2006) who also found increased cancer-related mortality in T2DM patients treated with insulin or sulfonylureas. In the 2010 multinational case–control study by Maisonneuve et al. (2010), a history of DM conveyed an increased RR for PC overall (2.16), which was 6.68 for those whose history of DM started within 1 year of the diagnosis of cancer, falling to 1.28 for those whose diabetic history had existed for more than 10 years. In addition, the type of treatment for the DM was found to have a differential effect, with those who had been treated with insulin having an RR of 3.54, while those who had been treated only with oral agents having an RR of 1.78. These findings confirmed those of Bonelli et al. who similarly found a higher risk associated with insulin treatment than with non-insulin therapy. The higher risk seen with insulin treatment in the case–control studies of Maisonneuve et al. and Bonelli et al. may reflect a direct effect of insulin on tumor cell proliferation as insulin therapy would also be expected to produce periods of hyperinsulinemia. In addition, however, other factors may have contributed to the results, including length of treatment and the degree of hyperglycemia present, both of which would be expected to be greater in insulin-requiring patients. Carstensen et al. (2012) recently analyzed the effect of diabetes and diabetic therapy on cancer incidence in the Danish health system over a 27-year period and noted that the adverse effects of insulin treatment were quite modest. Therefore, caution has been advised in the acceptance of a cancer-promoting effect of insulin until further data are available (Gerstein 2010, Giovannucci et al. 2010, Hernandez-Diaz & Adami 2010).
Insulin therapy has also been associated with a greater than expected incidence of breast cancer and other malignancies (Home & Lagarenne 2009, Jonasson et al. 2009), however, and insulin secretagogues have similarly been implicated in an increased incidence of PC. Case–control studies indicate an increased risk of cancer overall in T2DM patients treated with sulfonylureas (Bowker et al. 2006, Monami et al. 2009), and studies of PC in particular indicate that sulfonylurea therapy confers an increased risk (Currie et al. 2009, Li et al. 2009).
Incretin-based therapies (glucagon-like peptide-1 (GLP1) analogues and inhibitors of didpeptidyl peptidase-IV, the enzyme that metabolizes GLP1) are the newest form of insulin secretagogue treatment of T2DM. No long-term epidemiological studies have yet been reported of their effect on the incidence of PC due to the relatively short period of clinical availability of these agents. However, laboratory studies suggest an adverse effect of these agents on the development of pre-neoplastic or malignant lesions of the pancreas, possibly related to the growth-promoting effects of GLP1 and its analog exendin-4. Matveyenko et al. (2009) described the effects of sitagliptin, a DDP-IV inhibitor, on pancreatic ductal cell proliferation and β-cell mass in the human islet amyloid polypeptide (HIP) transgenic model of T2DM in rats. Sitagliptin was seen to induce β-cell replication, β-cell apoptosis, pancreatic ductal metaplasia, and a fourfold increase in duct cell proliferation, all effects that were blocked by concurrent metformin administration (Fig. 5).
Gier et al. (2012) recently reported that the GLP1 analog exendin-4 increased duct cell replication and increased the development of dysplastic pancreatic intraepithelial neoplasia (PanIN) lesions in a rat model in which activated KrasG12D was induced. These investigators found that pancreatic duct glands (thought to be precursors of PanINs) in both rats and humans contain GLP1 receptors and are increased by treatment with exendin-4.
GLP1 and its analogs have been shown to increase expression of the transcription factor pancreatic–duodenal homeobox-1 (PDX1; Perfetti et al. 2000, Stoffers et al. 2000). PDX1 is critical for normal pancreatic development and has been shown to be overexpressed in PC (Koizumi et al. 2003, Wang et al. 2005). Furthermore, PDX1 increases cell proliferation, invasion, and colony formation of transformed cell lines in vitro and stimulates PC formation in severe combined immunodeficiency (SCID) mice, thereby fulfilling criteria as an oncogene (Liu et al. 2011). The gene NR5A2 (also known as LRH1) has been identified as a PC susceptibility gene (Petersen et al. 2010) and is a target gene of PDX1 expression (Annicotte et al. 2003). PDX1-expressing endocrine cells can be transformed into a malignant, ductal cell phenotype after induction of pancreatitis by cerulein in mice (Gidekel Friedlander et al. 2009), which supports the epidemiological findings of a striking association of CP with the development of PC (Greer & Whitcomb 2009). These findings suggest that Kras activation, either by genetic or nongenetic (inflammatory) events, may therefore activate neoplastic transformation within the pancreas to PDX1-expressing islet and/or ductal cells. Agents that increase PDX1 expression, such as GLP1 and its analogs, may therefore facilitate this process.
Most recently, Elashoff et al. (2011) examined the database of adverse outcomes reported to the US Food and Drug Administration and discovered that compared with other antidiabetic therapies, the use of sitagliptin or the GLP1 analog exenatide was associated with a two- to three-fold increase in the reporting incidence of PC. Although subject to multiple methodological limitations, these findings nevertheless raise concerns regarding an increased risk for pancreatitis and PC with GLP1-based therapy. Further epidemiological studies are clearly needed to define the possible hazards of incretin-based antidiabetic therapy, particularly in patients with T3cDM due to CP (Cui & Andersen 2011).
Evidence for decreased PC incidence in patients treated with metformin
Although insulin and insulin secretagogue therapy is associated with an increased risk of PC, numerous clinical studies have demonstrated that the administration of metformin in DM patients exhibits a protective effect that is manifested by a decreasing incidence of different tumors and an improved prognosis of patients with cancer. Metformin has been widely used for the treatment of T2DM for almost 50 years, and its effectiveness has been attributed to enhanced sensitivity to insulin, rather than to an insulinotropic action. Its safety and efficacy are so well established that it is recommended as the first line of therapy for T2DM (Nathan et al. 2009). Evans et al. (2005) evaluated metformin use in diabetics who were admitted to hospital with a diagnosis of cancer between 1993 and 2001 in Tayside, Scotland, and compared this cohort with diabetic controls who were not admitted for cancer. They found that any exposure to metformin was associated with a significant reduction in cancer risk (RR=0.77). The same group subsequently examined the database of all diabetic patients in the region and compared outcomes based on the national registry of cancer deaths (Libby et al. 2009). When the outcome of 4804 metformin users was compared with 4085 nonusers, a reduced risk (RR 0.63) for cancer mortality was found among diabetic patients treated with metformin, whereas a (nonsignificantly) increased risk of cancer mortality was seen among insulin and sulfonylurea users. Li et al. (2009) analyzed 978 patients with PC, including 259 diabetics, and 863 controls, including 109 diabetics, to assess the effects of diabetes treatment on the risk of PC. They found that whereas insulin treatment was associated with an increased risk (RR 2.78 for insulin users of more than 5 years), metformin therapy was associated with a 70% decreased risk (RR 0.30) for similar long-term users of the drug. The implications are further supported by a more recent cohort study from Taiwan in which metformin users were found to have a 85% decreased risk of pancreatic malignancy (HR 0.15, 95% CI: 0.03–0.79; Lee et al. 2011b). Sadeghi et al. (2011) performed a retrospective cohort study of 302 patients to investigate the survival benefit of metformin in patients with DM and pancreatic malignancy. They report the median survival to be longer in metformin users when compared with nonusers: 16.6 vs 11.5 months (P=0.0044). They also report a 33% decreased risk of death in patients who used metformin compared with those who did not. In addition to studies of its anticancer effects in PC, metformin has also been shown to have anticancer effectiveness in diabetic breast cancer patients (Jiralerspong et al. 2009, Chlebowski et al. 2012) and in colon cancer patients (Zhang et al. 2011).
These clinical studies corroborated the initial findings of Schneider et al. (2001) who found a therapeutic effect of metformin in a hamster model of carcinogen-induced PC. In these animals, metformin treatment significantly decreased islet cell hyperplasia and pancreatic ductal proliferation and completely prevented the development of pancreatic adenocarcinoma. Additional evidence showed that metformin can inhibit PC cell growth and proliferation by disrupting the cross talk between insulin/IGF1Rs and G-protein-coupled receptors through the activation of the liver kinase B1–AMP (LKB1–AMP) AMPK pathway, which serves not only to suppress hepatic glucose production and reduce the need for insulin but also inhibits the signaling mechanisms that regulate cellular proliferation (Rozengurt et al. 2010). LKB1 is a known tumor suppressor that activates AMPK, a potent inhibitor of mTOR complex 1, which serves as a regulator of protein synthesis and replication (see Fig. 4). Bao et al. (2012) found that metformin significantly decreased cell survival, clonogenicity, wound healing capacity, sphere-forming capacity (pancreatospheres), and increased disintegration of pancreatospheres in both gemcitabine-sensitive and gemcitabine-resistant PC cells. Metformin also decreased the expression of cancer stem cell (CSC) markers, CD44, EpCAM, EZH2, Notch1, Nanog, and Oct4 and caused re-expression of miRNAs (let-7a, b, miR-26a, miR-101, and miR-200b,c) that are typically lost in PC. These results suggest that the biological effects of metformin as an antineoplastic agent are mediated through re-expression of miRNAs and decreased expression of CSC-specific genes that are affected by carcinogenesis, and numerous laboratory and clinical studies are now underway to examine the antitumor applications of metformin.
Diabetes as a consequence (and harbinger) of PC
The prevalence of diabetes is increased in PC patients
Many cohort and case–control studies indicate that 25–50% of patients with a diagnosis of PC will have developed DM within 1–3 years before their diagnosis of malignancy (Huxley et al. 2005, Chari et al. 2008). This implies that recent-onset DM associated with PC is caused by pancreatic malignancy and suggests that DM is a biomarker of early-stage PC. The problem is that new-onset DM, per se, is not a powerful-enough predictor of PC to stand alone as an indication for radiological or endoscopic screening, as 98% of patients with adult-onset DM will never develop PC (Chari et al. 2005a). Imaging protocols applied to patients with new-onset DM older than the age of 50 have not been shown to be either practical or reliable as an early detection method based on the studies in Japan (Ogawa et al. 2002), France (Damiano et al. 2004), and the USA (Chari et al. 2005a,b).
In addition to the frequent development of DM just proximate to the diagnosis of PC, the concept that PC is a cause of new-onset DM is supported by several observations. Patients with premalignant pancreatic lesions from kindreds in which PC is highly prevalent typically have concurrent DM (Brentnall et al. 1999). Furthermore, new-onset DM associated with PC has been seen to resolve after the successful resection of the tumor (Permert et al. 1993, Fogar et al. 1994, Pannala et al. 2008; Fig. 6). In laboratory studies, PC-derived cell lines induce hyperglycemia in SCID mice (Basso et al. 1995), and a PC-derived S-100A8 N-terminal peptide has been identified as a diabetogenic agent (Basso et al. 2006).
Despite vigorous clinical investigation, no biomarker specific for PC-associated DM has been validated, and tumor markers such as carcinoembryonic antigen and carbohydrate antigen 19-9 are insufficiently sensitive to detect early-stage disease. Permert et al. (1994) identified elevated levels of islet amyloid polypeptide (IAPP), a β-cell peptide co-secreted with insulin, as a possible biomarker of PC and demonstrated that PC cells are capable of inducing IAPP release from β-cells (Ding et al. 1998), but Chari et al. (2001) subsequently showed that IAPP was not sufficiently sensitive to serve as a biomarker of PC. Using microarray analysis, Pfeffer et al. (2004) identified connexin 26, a gap junction protein, as being highly overexpressed in PC patients with DM. Using similar methods, Huang et al. (2010) identified two upregulated genes in 27 patients with PC-associated DM, vanin-1 and matrix metalloproteinase 9, that together showed the best correlation with the diagnosis. No large-scale studies have been reported yet regarding the sensitivity and specificity of these possible biomarkers.
Pathogenesis of pancreatic carcinoma-associated diabetes
The mechanism(s) responsible for the DM caused by PC are incompletely understood. Defects in insulin sensitivity and insulin secretory capacity have been identified in patients with PC-associated DM, and abnormalities in glucose metabolism in skeletal muscle and liver have been observed in vitro as well. These multiple abnormalities suggest that one or more humoral factors are likely involved in PC-associated DM, as reviewed by Pannala et al. (2009).
Animal models of carcinogen-induced PC exhibit both hyperinsulinemia (Liu et al. 2000) and insulin secretory impairments (Ahren & Andren-Sandberg 1993, Permert et al. 2001), which implies that insulin resistance is an early event in PC-associated DM. Impaired sensitivity to insulin has been demonstrated in euglycemic glucose clamp studies and by means of the homeostasis model assessment method in patients with PC-associated DM (Cersosimo et al. 1991, Chari et al. 2005a), which improves after removal of the tumor despite reduced insulin secretory capacity after surgical resection (Permert et al. 1993, Pannala et al. 2008). Clinical studies on the pathophysiology of PC-associated DM have been confounded, however, by the inability of investigators to document whether the DM exhibited by each PC patient is preexisting T2DM or pancreatogenic (type 3c) DM caused by the tumor.
Distinguishing T3cDM from T2DM
Recalling that T3cDM is the form of DM that is a consequence of pancreatic exocrine pathology and that T3cDM is in many ways distinct from T2DM (see Table 1), the issue therefore becomes how to feasibly distinguish individuals with T3cDM from the vastly more prevalent T2DM, so as to identify the subset of high-risk patients in whom high-resolution imaging tests might feasibly identify suspicious areas within the pancreatic parenchyma.
A retrospective cohort study of 2122 diabetic patients suggested that PC developed within 3 years after the diagnosis of DM in 1% of the patients who were at least 50-year old (Chari et al. 2005a). Johnson et al. (2011) found that the diagnosis of PC was highest within 3 months following the onset of DM, and Aggarwal et al. (2012) showed that the duration of DM before the diagnosis of PC by primary care providers averaged 6.5 months. A study by Lee et al. (2011a) reported that compared with the control group, PC patients were, on average, older, had more weight loss, lower usual BMI, a greater family history of PC (3.3 vs 0.7%; P=0.044), and had a lower family history of DM (13.9 vs 37.4%; P<0.001). These authors concluded that PC-associated DM (T3cDM) could be discriminated from new-onset T2DM based on clinical features, such as the lack of a family history of DM, age 65 years or older, recent weight loss of >2 kg, or a premorbid or usual BMI <25 kg/m2. A definitive diagnosis of T3cDM, however, requires further testing for confirmation in patients who lack a history of pancreatic disease or a family history of PC. Fecal elastase-1 levels <100 μg/g strongly suggest pancreatic exocrine impairments (Loser et al. 1996), but the most consistent laboratory finding in T3cDM due to any cause is a deficiency of the pancreatic polypeptide (PP) response to ingested nutrients (Cui & Andersen 2011, Magruder et al. 2011).
PP deficiency: a marker of T3cDM
PP is localized predominantly to islets in the ventral portion (head) of the pancreas and is promptly secreted in response to ingested nutrients. PP regulates the expression and availability of hepatic IRs, and hepatic insulin resistance due to PP deficiency is reversed by PP administration in animals and man (Andersen 2007). T3cDM secondary to cystic fibrosis, CP, pancreatic malignancy, or pancreatic resection is uniformly characterized by a deficiency in the nutrient-stimulated release of PP (Fig. 7) and a defect in hepatic insulin sensitivity (Slezak & Andersen 2001, Cui & Andersen 2011). T2DM, on the other hand, is typically associated with an increase in basal and nutrient-stimulated levels of PP (Glaser et al. 1988). Furthermore, healthy elderly subjects with normal glucose tolerance also demonstrate elevations in basal and nutrient-stimulated levels of PP compared with younger subjects (Magruder et al. 2011; Fig. 8). Therefore, the discrimination of T3cDM from T2DM is based on the failure of plasma PP levels to increase after nutrient ingestion. Basal levels of PP in PP-deficient subjects are similar to basal levels in normal subjects, so a nutrient stimulus is required to confirm PP deficiency. Glucose ingestion is a relatively weak stimulus for PP release, whereas a mixed nutrient meal is a strong inducer of PP release. A standardized mixed-nutrient stimulus is 8 ounces of a liquid dietary supplement such as Ensure-Plus; peak levels of PP are seen within 30–60 min after ingestion. Studies are currently in progress to establish the prevalence of PP deficiency in PC patients with and without DM, but a diagnosis of new-onset T3cDM based on PP deficiency is a strong indicator that the patient is in a high-risk category for PC and warrants further investigation to identify a pancreatic parenchymal abnormality.
Prevalence of T3cDM
Patients who have a known history of pancreatic disease and who are also diabetic require special surveillance for PC. The most common cause of T3cDM is CP, which harbors a 10- to 20-fold increased risk of PC; the combination of CP and DM increases this risk 33-fold. Unless drug side effects (principally gastrointestinal sensitivity) or a contraindication (such as renal insufficiency) is present, metformin should be prescribed to all T3cDM patients, even if other antidiabetic drugs such as insulin have to be added for adequate glycemic control (Cui & Andersen 2011).
T3cDM had been estimated to account for only 1–2% of all diabetic patients in North America (Ganda 2005) but is known to affect as many as 15–20% of diabetic patients in the Indian and Southeast Asian continents, where tropical or fibrocalcific pancreatitis is endemic (Abu-Bakare et al. 1986). In a careful evaluation of almost 2000 diabetic patients referred to an academic medical center in Germany, Hardt et al. (2008) discovered that 8% of all diabetic patients harbored T3cDM (Fig. 9). Furthermore, Ewald et al. (2011) subsequently showed that nearly half of the T3cDM patients had been previously misdiagnosed as either T1DM (6%) or T2DM (40%). With the advent of improved imaging methods to detect pancreatic pathology, and the availability of a practical screening method to quantify exocrine pancreatic function, previous estimates of the prevalence of T3cDM are now understood to have been spuriously low (Ewald et al. 2009).
In the German cohort study, 78.5% of T3cDM patients had CP as the underlying etiology of their T3cDM. In an older study of 500 patients with CP due to alcoholism, DM developed in 83% within 25 years of the clinical onset of CP, and more than half of the diabetic patients ultimately required insulin therapy (Malka et al. 2000). With the increasing prevalence of CP worldwide (Rothenbacher et al. 2005), the population of patients at greatest risk for the development of PC is clearly expanding. Therefore, an aggressive approach to the identification, surveillance, and management of patients at high-risk for PC is an important strategy in order to reduce the mortality rate of PC.
Drugs used to treat PC may cause diabetes
A recently emerging issue is the possible adverse effect on glucose metabolism of anticancer therapies. Cancer patients commonly exhibit hyperglycemic states or DM following glucocorticoid administration (Saylor & Smith 2009). The increasing use of targeted chemotherapy directed against components of the IGF1 pathway may amplify the frequency of anticancer drug-related diabetes. IGF1 and insulin, their receptors, and their intracellular signaling pathways share multiple similarities. Likewise, the biological (metabolic and mitogenic) effects of insulin and IGFs overlap. Hyperglycemia was observed in some patients enrolled in studies with an anti-IGF1R antibody (Haluska et al. 2006, Lacy et al. 2008). This is likely to be a consequence of a compensatory increase in the circulating concentration of GH after IGF1 blockade, with the consequent increase in GH-induced insulin resistance (del Rincon et al. 2007). Hyperglycemia, hypertriglyceridemia, and hypercholesterolemia were also observed in about 20% of patients treated with the mTOR inhibitors (Bellmunt et al. 2008). Recent reports also documented increased blood glucose levels in 26% of temsirolimus-treated patients (Bellmunt et al. 2008, Malizzia & Hsu 2008). However, tyrosine kinase inhibitor therapy directed at IGF1Rs was associated with less hyperglycemia than IGF1R-blocking antibodies (Pollak 2008). At present, insufficient data are available to assess the possible diabetogenic effects of phosphoinositol-3-kinase and AKT inhibitor therapy.
Conclusions and clinical recommendations
Epidemiological data clearly demonstrate an etiological link between long-standing T2DM, and probably T1DM, and PC. Successful treatment of T2DM and obesity has been shown to reduce the risk of PC, but treatment with insulin, insulin analogs, and insulin secretagogues increases or maintains the risk. Metformin has been shown to reduce the risk of PC owing to its antidiabetic and antineoplastic actions and should be considered as first-line therapy in all new-onset DM patients over the age of 50. In addition, metformin, as well as lifestyle alterations, have been shown to reduce the incidence of T2DM when administered to obese individuals with impaired glucose tolerance (Knowler et al. 2002, Cefalu 2012). Such successful prevention strategies need to be more fully used in order to reverse the increasing incidence of PC.
New-onset DM can be a consequence, and therefore a harbinger, of PC. DM caused by PC is classified as T3cDM, which occurs in up to 30% of patients with PC. If a new-onset DM patient does not have family history of DM, is aged 65 years or older, or has had a stable BMI <25 kg/m2, PC-associated DM (T3cDM) should be considered, especially within 3 months following onset of DM. If the patient has a family history of two or more relatives with PC, high-resolution imaging (preferably computer tomographic scanning followed by endoscopic ultrasound) of the pancreas should be considered to identify suspicious areas within the parenchyma (Brentnall et al. 1999).
If no family history of PC is present, T3cDM can be confirmed by documenting the absence of a rise in PP levels after a test meal or liquid mixed-nutrient challenge. Patients with confirmed T3cDM, with or without a history of pancreatic disease, are candidates for periodic high-resolution pancreatic imaging to detect early pathological changes within the pancreas. With an aggressive approach to distinguishing T3cDM from T2DM, many cases of early PC may be able to be identified when curative removal of the tumor is still possible.
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the review reported.
This review did not receive any specific grant from any funding agency in the public, commercial or not-for-profit sector.
ArslanAAHelzlsouerKJKooperbergCShuXOSteplowskiEBueno-de-MesquitaHBFuchsCSGrossMDJacobsEJLacroixAZ2010Anthropometric measures, body mass index, and pancreatic cancer: a pooled analysis from the Pancreatic Cancer Cohort Consortium (PanScan). Archives of Internal Medicine170791–802. (doi:10.1001/archinternmed.2010.63).
BaoBWangZAliSAhmadAAzmiASSarkarSHBanerjeeSKongDLiYThakurS2012Metformin inhibits cell proliferation, migration and invasion by attenuating CSC function mediated by deregulating miRNAs in pancreatic cancer cells. Cancer Prevention Research5355–364. (doi:10.1158/1940-6207.CAPR-11-0299).
DingXZFehsenfeldDMMurphyLOPermertJAdrianTE2000Physiological concentrations of insulin augment pancreatic cancer cell proliferation and glucose utilization by activating MAP kinase, PI3 kinase and enhancing GLUT-1 expression. Pancreas21310–320. (doi:10.1097/00006676-200010000-00014).
EdwardsBKBrownMLWingoPAHoweHLWardERiesLASchragDJamisonPMJemalAWuXC2005Annual report to the nation on the status of cancer, 1975–2002, featuring population-based trends in cancer treatment. Journal of the National Cancer Institute971407–1427. (doi:10.1093/jnci/dji289).
Ganda OP 2005 Secondary forms of diabetes. In Joslin's Diabetes Mellitus 14th edn pp 477–492. Eds CR Kahn GC Weir GL King AM Jacobson AC Moses & RJ Smith. Philadelphia PA USA: Lippincott Williams and Wilkins
GierBMatveyenkoAVKirakossianDDawsonDDrySMButlerPC2012Chronic GLP-1 receptor activation by exendin-4 induces expansion of pancreatic duct glands in rats and accelerates formation of dysplastic lesions and chronic pancreatitis in the KrasG12D mouse model. Diabetes611250–1262. (doi:10.2337/db11-1109).
HaluskaPCarboniJMLoegeringDALeeFYWittmanMSaulnierMGFrennessonDBKalliKRConoverCAAttarRM2006In vitro and in vivo antitumor effects of the dual insulin-like growth factor-I/insulin receptor inhibitor, BMS-554417. Cancer research66362–371. (doi:10.1158/0008-5472.CAN-05-1107).
JiaoLBerrington de GonzalezAHartgePPfeifferRMParkYFreedmanDMGailMHAlavanjaMCAlbanesDBeane FreemanLE2010Body mass index, effect modifiers, and risk of pancreatic cancer: a pooled study of seven prospective cohorts. Cancer Causes & Control211305–1314. (doi:10.1007/s10552-010-9558-x).
JiralerspongSPallaSLGiordanoSHMeric-BernstamFLiedtkeCBarnettCMHsuLHungMCHortobagyiGNGonzalez-AnguloAM2009Metformin and pathologic complete responses to neoadjuvant chemotherapy in diabetic patients with breast cancer. Journal of Clinical Oncology273297–3302. (doi:10.1200/JCO.2009.19.6410).
JohansenDStocksTJonssonHLindkvistBBjörgeTConcinHAlmquistMHäggströmCEngelandAUlmerH2010Metabolic factors and the risk of pancreatic cancer: a prospective analysis of almost 580,000 men and women in the Metabolic Syndrome and Cancer Project. Cancer Epidemiology Biomarkers & Prevention192307–2317. (doi:10.1158/1055-9965.EPI-10-0234).
KnowlerWCBarrett-ConnorEFowlerSEHammanRFLachinJMWalkerEANathanDM2002Diabetes program prevention research group. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. New England Journal of Medicine346393–403. (doi:10.1056/NEJMoa012512).
LacyMQAlsinaMFonsecaRPaccagnellaMLMelvinCLYinDSharmaAEnriquez SaranoMPollakMJagannathS2008Phase I, pharmacokinetic and pharmacodynamic study of the anti-insulin like growth factor type 1 Receptor monoclonal antibody CP-751,871 in patients with multiple myeloma. Journal of Clinical Oncology263196–3203. (doi:10.1200/JCO.2007.15.9319).
LeeMSHsuCCWahlqvistMTsaiHNChangYHHuangYC2011bType 2 diabetes increases and metformin reduces total, colorectal, liver, and pancreatic incidences in Taiwanese: a representative population prospective cohort study of 800,000 individuals. BMC Cancer1120. (doi:10.1186/1471-2407-11-20).
LowenfelsABMaisonneuvePCavalliniGAmmannRWLankischPGAndersenJRDimagnoEPAndrén-SandbergADomellöfL1993Pancreatitis and the risk of pancreatic cancer. International Pancreatitis Study Group. New England Journal of Medicine3281433–1437. (doi:10.1056/NEJM199305203282001).
MaisonneuvePLowenfelsABBueno-de-MesquitaHBGhadirianPBaghurstPAZatonskiWAMillerABDuellEJBoffettaPBoyleP2010Past medical history and pancreatic cancer risk: results from a multicenter case–control study. Annals of Epidemiology2092–98. (doi:10.1016/j.annepidem.2009.11.010).
MatveyenkoAVDrySCoxHIMoshtaghianAGurloTGalassoRButlerAEButlerPC2009Beneficial endocrine but adverse exocrine effects of sitagliptin in the human islet amyloid polypeptide transgenic rat model of type 2 diabetes: interactions with metformin. Diabetes581604–1615. (doi:10.2337/db09-0058).
NathanDMHolmanRRBuseJBSherwinRDavidsonMBZinmanBFerranniniE2009Medical management of hyperglycemia in type 2 diabetes: a consensus algorithm for the initiation and adjustment of therapy. A consensus statement by the American Diabetes Association and the European Association for the Study of Diabetes. Diabetes Care32193–203. (doi:10.2337/dc08-9025).
PerfettiRZhouJDoyleMEEganJM2000Glucagon-like peptide-1 induces cell proliferation and pancreatic-duodenum homeobox-1 expression and increases endocrine cell mass in the pancreas of old, glucose intolerant rats. Endocrinology1414600–4605. (doi:10.1210/en.141.12.4600).
PetersenGMAmundadottirLFuchsCSKraftPStolzenberg-SolomonRZJacobsKBArslanAABueno-de-MesquitaHBGallingerSGrossM2010A genome-wide association study identifies pancreatic cancer susceptibility loci on chromosomes 13q22.1, 1q32.1 and 5p15.33. Nature Genetics42224–228. (doi:10.1038/ng.522).
PfefferFKoczanDAdamUBenzSvon DobschuetzEPrallFNizzeHThiesenHJHoptULoblerM2004Expression of connexin 26 in islets of Langerhans is associated with impaired glucose tolerance in patients with pancreatic carcinoma. Pancreas29284–290. (doi:10.1097/00006676-200411000-00007).
del RinconJPIidaKGaylinnBDMcCurdyCELeitnerJWBarbourLAKopchickJJFriedmanJEDrazninBThornerMO2007Growth hormone regulation of p85alpha expression and phosphoinositide 3-kinase activity in adipose tissue: mechanism for growth hormone mediated insulin resistance. Diabetes561638–1646. (doi:10.2337/db06-0299).
RozengurtESinnett-SmithJKisfalviK2010Crosstalk between insulin/insulin-like growth factor-1 receptors and G protein-coupled receptor signaling systems: a novel target for the antidiabetic drug metformin in pancreatic cancer. Clinical Cancer Research: an Official Journal of the American Association for Cancer Research162505–2511. (doi:10.1158/1078-0432.CCR-09-2229).
(Y Cui and D K Andersen contributed equally to this work)