The Associations of Advanced Glycation End Products and Its Soluble Receptor with Pancreatic Cancer Risk: A Case–Control Study within the Prospective EPIC Cohort

Advanced glycation end products (AGE) are formed by nonenzymatic reactions of reduced sugars, such as glucose, with amino groups in proteins, lipids, and nucleic acids. Exogenous sources of AGE are those derived from tobacco and those from thermally processed food, so called dietary AGEs, whereas endogenously, AGEs can be formed in a wide range of body tissues and cell types. The endogenous formation of AGE is slow under nonpathologic conditions but enhanced during hyperglycemia. One of the most prominent and also best studied AGE is Nϵ-(carboxymethyl)lysine (CML), a glycoxidation product (1, 2). AGEs are thought to exert their proinflammatory effects by binding to receptors, with RAGE being the best characterized (2, 3). In blood, soluble forms of RAGE have been detected, including splice variants such as the most prominent endogenous secreted RAGE (esRAGE or RAGEv1; ref. 4) or cleavage forms of membrane-bound full-length RAGE (5). These soluble forms of RAGE are suspected to bind free AGEs and, therefore, might act as “decoy receptors,” preventing RAGE ligands from interaction with cell surface RAGE. Consequently, this is thought to inhibit angiogenesis and tumor cell activation (6).

AGE and RAGE are expressed in many tissues and cell types (2, 7), and for many years, both have been implicated in a number of metabolic, neurodegenerative, and inflammatory diseases such as diabetes mellitus and vascular diseases (8). More recently, involvement of the AGE/RAGE axis in cancer has been suspected (6, 7, 9, 10). Human investigations on AGE and RAGE concentrations and their associations with pancreatic cancer risk are, however, limited. One hospital-based case–control study has shown decreased levels of soluble RAGE among their 51 patients with pancreatic cancer compared with cancer-free control subjects (11), and a prospective case–control study among male Finnish smokers with 255 cases observed a decrease in risk of pancreatic cancer with elevated soluble RAGE, whereas CML levels were not related to pancreatic cancer risk (12). Up to date, no prospective study has been conducted investigating the relationship of the splice variant esRAGE with the risk of pancreatic cancer.

To elucidate the suspected positive relationship of CML and the suspected inverse relationship of esRAGE with risk of pancreatic cancer, we conducted a nested case–control study within the European Prospective Investigation into Cancer and Nutrition (EPIC), using data from 454 pancreatic cancer subjects and the equal number of cancer-free control subjects selected from the general Western European population.

EPIC is a large prospective cohort study with 519,978 participants enrolled between 1992 and 2000 in 23 centers across 10 European countries (Denmark, France, Germany, Greece, Italy, the Netherlands, Norway, Spain, Sweden, and the United Kingdom). At baseline, blood was taken and detailed questionnaires were distributed, covering among others habitual diet, anthropometric measures, data on physical activity, socioeconomic status, smoking, and medical conditions such as history of diabetes. Study design, population, and baseline data collection have been described previously in detail (13, 14). EPIC as well as this individual project were approved by the local ethical review committees, and each study participant provided informed consent.

In all EPIC centers, data on vital status are collected through mortality registries, in combination with health insurance data (France) or active follow-up (Greece). Population cancer registries (Denmark, Italy, the Netherlands, Spain, Sweden, and the United Kingdom) or a combination of methods including insurance records, cancer and pathology registries, and active follow-up through study subjects (France, Germany, Greece) were used to identify incident cancer cases. For the present project on pancreatic cancer, closure date of the study period were defined as the latest date of complete follow-up for both cancer incidence and vital status in each EPIC center; varying from December 2002 to December 2005. The last known contact, the date of diagnosis, or the date of death, whichever came first, were defined as the end of follow-up in Germany, Greece, and France.

Exocrine pancreatic cancer incident data were coded as C25 (25.0–25.3, 25.7–25.9) according to ICD-10. Cases were selected between both sexes who developed pancreatic cancer after their recruitment into the study and before the end of the study period. Exclusion criteria were occurrence of other malignant tumors preceding pancreatic cancer diagnosis, except nonmelanoma skin cancer and nonavailability of blood specimens. Follow-up has led to the identification of 578 primary exocrine pancreatic adenocarcinomas, for 466 of these blood specimens were available, and for 454 sRAGE and CML could be measured. Most tumors occurred in the pancreatic head (42%) with few in the body (7%) and tail (5%), whereas the rest of the tumors were of unknown localization. Three hundred and thirty-three (73%) of the pancreatic cancer cases were microscopically confirmed and the remaining 34% were diagnosed by physical examination, imaging results, or clinical symptoms. For each case, one control alive and free of cancer at the time of diagnosis of the case was selected using an incidence density sampling procedure. Cases and controls were individually matched, using the following matching criteria: center, sex, age at blood collection (±3 years), date of blood donation (±3 months), time of blood donation (±2 hours), fasting status (<3, 3–6, >6 hours after last meal) and, in women, use of hormones (oral contraceptive pill, hormone, or estrogen replacement therapy).

Blood samples of roughly 420,000 EPIC participants were aliquoted and either stored in liquid nitrogen (−196°C) at a central biorepository or locally in freezers at −70°C (Sweden) or nitrogen vapor (−150°C; Denmark). Serum samples (and EDTA plasma for the Swedish center Umeå) from one center were analyzed within the same analytic batch in the specialized immunoassay laboratory of the Division of Cancer Epidemiology (Heidelberg, Germany). ELISA was used to measure esRAGE in serum and plasma samples (B-Bridge International, Inc.). Measurements for CML were conducted with ELISA from a different company in serum samples only, after enzymatic pretreatment with Proteinase K (Synvista Therapeutics, Inc.). Therefore, plasma samples from Umeå were not included in the CML analysis. In total, esRAGE could be measured for 886 subjects and CML for 832 subjects. Intra- and interbatch coefficients of variation were 2.2% and 5.2% for esRAGE and 10.6% and 19.4% for CML. Units for esRAGE are expressed as picogram per milliliter and for CML as nanogram per milliliter.

Partial Spearman rank correlation coefficients (ρ) adjusted for age, sex, and EPIC recruitment center were used to assess the correlations of CML and esRAGE with anthropometric measures, glycated hemoglobin (HbA1c), age, and other variables of interest such as dietary factors or inflammatory markers. HbA1c and inflammatory markers have been measured previously in the same study population, and the results on HbA1c have been published recently (15).

Conditional logistic regression models were applied to estimate ORs and 95% confidence intervals (CI) for the associations of CML and esRAGE with pancreatic cancer risk, by either modeling the exposure variables continuously or using quintiles. Cutoff points for the categories were based on the distribution of controls and trend tests over quintiles were assessed by modeling the median value within each category as a continuous variable. Continuous measurements of CML and esRAGE were log-transformed to achieve approximate normality. To assess the association of bioavailable CML with risk of pancreatic cancer, we calculated the CML/esRAGE ratio and used this variable in all statistical models in addition to CML and esRAGE.

Conditional logistic regression was also used to investigate the relationship of potential confounders with cancer risk, including body mass index (BMI), waist circumference, waist–hip ratio, dietary factors (daily average intakes of alcohol, total fat, total protein, carbohydrates, glucose, red meat, processed meat), smoking status (never, former, current, unknown), physical activity [Cambridge index (ref. 16): active, moderately active, moderately inactive, inactive], and diabetes. Subjects were classified as diabetics in the current study if they had baseline HbA1c levels ≥6.5% and/or self-reported diabetes at recruitment (n = 93). Variables were considered as confounders if they were associated with the exposure and the outcome and if they changed the logistic β-estimate in a multivariate model by more than 10%. We finally adjusted for BMI as a continuous variable and for smoking status categorically, with the following categories: never-smoker; former smoker who stopped less than 10 years ago, former smoker who stopped 10 or more years ago; current smoker with 1–9, 10–19, or ≥ 20 cigarettes per day; and smoking status unknown. In addition, logistic regression models were mutually adjusted for esRAGE and CML. HbA1c was added in further models to explore the potential additional confounding effect of progressively deteriorating glucose tolerance and of diabetes and to investigate the effect of hyperglycemia as an enhancer of AGE formation.

In addition, analyses were stratified by factors that could modify the relationship between CML and esRAGE and pancreatic cancer, such as diabetes. Heterogeneity of effect was assessed by adding cross-product terms into the logistic regression models over continuous levels of CML and esRAGE and testing the significance with the Wald test, crude and adjusted for BMI and smoking.

All statistical analyses were conducted using the Statistical Analysis System (SAS) software package, Version 9.2 (SAS Institute Inc.). All statistical tests were two-tailed and threshold of significance was 0.05.

Pancreatic cancer cases were on average 63 years old and had a mean follow-up time of 5.3 years (range, 0–13 years; Table 1). At baseline, a higher proportion of cases reported being diabetic (14% vs. 8%) or current smokers (31% vs. 22%) than controls. No case–control differences by waist–hip ratio were observed among men and women, and no differences by BMI or waist circumference among men. Female pancreatic cancer cases had higher BMI and larger waist circumference than female controls.

CML and esRAGE correlated weakly but significantly with each other among controls (ρ = 0.15; 95% CI, 0.04–0.24), and both markers correlated negatively with anthropometric measures with partial correlation coefficients up to −0.27 (Table 2). Negative correlations were also observed for esRAGE and CML with inflammatory markers such as C-reactive protein (CRP; Table 2). Controls with diabetes had lower CML (geometric mean = 689 ng/mL; 95% CI, 625–760) and esRAGE levels (372 pg/mL; 95% CI, 321–432) than nondiabetic controls (737 pg/mL; 95% CI, 719–755 and 457 pg/mL; 95% CI, 443–437, respectively). CML levels were lowest among current smokers than nonsmokers but no differences in esRAGE levels were observed by smoking status (data not shown). CML and esRAGE levels correlated significantly with daily average intakes of glucose (Table 2) but not with intakes of total fat, total protein, red meat, processed meat, or alcohol use (data not shown).

Elevated CML levels tended to be associated with a reduced pancreatic cancer risk (crude OR = 0.57; 95% CI, 0.32–1.01, comparing highest vs. lowest quintile; P_trend = 0.05; Table 3). Adjustments for smoking, BMI, HbA1c, and esRAGE attenuated the association of CML with pancreatic cancer risk (OR = 0.78; 95% CI, 0.41–1.49; P_trend = 0.483), with BMI having the strongest effect. In contrast, esRAGE levels were not associated with pancreatic cancer risk (crude OR = 0.98; 95% CI, 0.62–1.54; comparing highest with lowest quintile). Additional adjustments for inflammatory markers or dietary factors had a negligible effect on the risk estimates for CML and esRAGE (data not shown). The CML/esRAGE ratio was not associated with risk of pancreatic cancer (Table 3).

The associations of elevated CML and esRAGE levels with pancreatic cancer risk differed statistically significantly by diabetes and smoking status. The association with esRAGE was also modified by follow-up time [time between recruitment (blood collection) and date of tumor diagnosis; Fig. 1]. In these analyses, never and former smokers seemed to have a lower and current smokers a higher risk of pancreatic cancer with elevated CML levels. For higher concentrations of both biomarkers, risk was stronger among diabetic than among nondiabetic participants. However, none of the risk estimates were significant in either crude or adjusted analyses. For higher esRAGE levels, pancreatic cancer risk was significantly lower by 54% among participants with less than 2 years of follow-up, which was slightly stronger after multivariate adjustments, whereas there was no association among those with a longer follow-up. No heterogeneity of effects was seen by measures of adiposity, by fasting status, or by median intake of dietary exposures such as red or processed meat (data not shown). The association between CML and pancreatic cancer risk did not differ by levels of esRAGE and vice versa (data not shown).

Prospectively, we observed a roughly 40% decrease in pancreatic cancer risk with elevated CML levels (>1,017 vs. <560 ng/L), which disappeared after multivariate adjustments for HbA1c levels, BMI, smoking status, and esRAGE concentrations. Elevated esRAGE levels and the CML/esRAGE ratio were not associated with risk of pancreatic cancer.

RAGE and its ligands have been linked to several diseases, including diabetes and its complications, chronic inflammatory diseases (8), and more recently, to cancer (6, 7, 9, 10). Activation of the multiligand/RAGE axis has been shown to perpetuate inflammation at the tumor microenvironment level, stimulate angiogenesis, and support invasion and metastasis (10). In particular, the RAGE ligand S100P and RAGE itself have been found to be overexpressed in pancreatic cancer cells or tissue (7). In contrast, overexpression of the splice variant esRAGE has been shown to inhibit expression of genes involved in tumor formation, cell invasion, and angiogenesis in various cells (17). However, data on CML and esRAGE levels in pancreatic cancer cells or plasma are sparse.

Ours is the third prospective study investigating the association of circulating levels of CML with risk of cancer. Jiao and colleagues did not detect any associations of CML levels with risk of colorectal cancer (18), but they observed a decrease in risk of pancreatic cancer in the unadjusted analyses, which became nonsignificant in the multivariable model (12). Our results are in line with the latter, with a borderline inverse association in the crude model. In contrast to our results of esRAGE, we did not observe any effect modification by follow-up time. On the basis of these results and the observation that plasma AGE levels not necessarily reflect tissue AGE levels (19), plasma levels of CML, and probably also of other AGE, might be elevated in tumor tissue only but not in plasma either many years prior to tumor diagnosis or shortly before its onset. Future studies are needed to clarify this hypothesis as, to date, no data are available on CML levels in pancreatic tumor tissue and only little is known on plasma circulating levels preceding cancer diagnosis.

The mechanisms by which esRAGE versus RAGE production is regulated are not fully understood (6). Secretion of esRAGE is a consequence of RAGE mRNA processing, possibly indicating enhanced RAGE expression through increased extracellular ligand binding or through intracellular activation of NF-κB secondary to cytokine release. Increasing esRAGE levels have been postulated to bind AGE and, therefore, prevent the activation of post-RAGE signaling (20). Targeted knockdown of RAGE in pancreatic tumor cells resulted in increased apoptosis, diminished autophagy, and decreased tumor cell survival whereas overexpression had the opposite effects (21). Lower sRAGE and esRAGE levels in tissue, in contrast, have been associated with various tumorigenic states (6). Whether plasma levels reflect tissue levels is not known. But if so, esRAGE (and sRAGE) levels should be decreased in blood of patients with pancreatic cancer. Indeed, this has been observed for sRAGE in one case–control study (11). In a prospective setting, this should translate into a decrease in pancreatic cancer risk with higher esRAGE levels, which we did not observe in our study. However, restricting the analyses to subjects developing pancreatic cancer within the first 2 years of follow-up (time between blood collection and tumor diagnosis), a risk decreasing effect was observed with higher esRAGE levels (OR = 0.46). Whether an undiagnosed tumor has led to lower esRAGE levels or whether lower esRAGE levels have been involved in pancreatic tumor development remains unanswered and needs to be addressed in further prospective studies with a large number of pancreatic cancer cases.

The overall association of pancreatic cancer risk with increasing CML and esRAGE levels seemed to be modified by the diabetes status of our subjects, that is, a rather strong increase in risk among diabetics and a null finding among nondiabetic participants. However, confidence intervals were wide and the test for heterogeneity was no longer significant after multivariate adjustments for BMI and smoking status. The first implicates low statistical power and the latter questions the observed effect modification by diabetes status. A recent population-based case–control study found higher blood levels of AGE and an increased RAGE expression but a decreased esRAGE expression and lower blood levels of esRAGE in diabetic than in nondiabetic participants (22). Our results are in line with the latter, that is, lower esRAGE levels in diabetics. We have no explanation for our findings of lower CML levels in diabetic participants, as this is in contrast to the above study and also in contrast to the biologic explanation of CML formation. After additional adjustment for HbA1c, ORs for pancreatic cancer risk with increasing CML and esRAGE levels were further off levels of significance in our study. On the basis of these observations, we suspect that elevated glucose levels and diabetes, but not altered CML and esRAGE levels, are the conditions associated with pancreatic cancer risk.

Our observed negative correlations of CML with anthropometric indices are in line with a recent publication (23). Semba and colleagues proposed biologic mechanisms for the correlations of CML with anthropometry, such that CML might be preferentially stored in fat tissue or metabolized in adipocytes. As overweight correlates with inflammatory markers, we suspected that the observed correlations of CML and esRAGE with inflammatory markers might be confounded by overweight, and, indeed, adjustments for BMI weakened these correlations. Therefore, the known link between overweight and inflammation might be the explanation for the observed correlations of CML with inflammatory markers.

Our study has several limitations. Drawing blood once does not necessarily reflect long-term levels of AGE products and its soluble receptor and repeated freeze thaw cycles such as in our study might affect CML and esRAGE concentrations in serum/plasma. Moreover, analyzing the agents in duplicate was not possible because of limited sample volume. Analyzing the matched case–control set within one assay, however, reduces to some extent laboratory measurement error. Matching on several variables bears the risk of overmatching and thus, may harm statistical efficiency, validity, and/or cost efficiency (24). However, the chosen variables are either standardization variables (age, sex), reflecting the variation in incidence of pancreatic cancer (study centre), or the differences in exposure levels at one point in time (fasting) and, therefore, are less likely to have introduced overmatching. If so, then rather an underestimation of the true risk may have occurred by using fasting status. A further limitation of our study is related to available questionnaire data and its impact on statistical analyses and interpretability of results. An inverse relationship has been observed between renal capacity and circulating CML and esRAGE concentrations in blood (25). We had no information on renal function or related diseases and, thus, could not control for this possible confounder. Our study is the largest with respect to the number of pancreatic cancer subjects in a prospective setting so far, analyzing the risk relationship of CML and esRAGE with primary exocrine adenocarcinoma of the pancreas. And it is the first to investigate esRAGE, as Jiao and colleagues analyzed sRAGE in their study of male Finnish smokers. sRAGE resembles membrane-bound full-length RAGE, esRAGE, and other splice variant forms of RAGE. We decided to measure esRAGE instead of sRAGE, as the first has been shown to be more stable over time than the latter and, in addition, only esRAGE was found to be capable of capturing AGE ligands but not other splice variants such as N-truncated RAGE (20). Investigating esRAGE in human cells and organs is a new evolving area of research and only little is known on its impact on diseases and disorders (6, 20, 26).

Future epidemiologic studies should analyze several AGE and RAGE derivatives in blood drawn several times, limit the number of freeze thaw cycles, and collect additional health- and disease-related information such as kidney function or diseases, diabetes, cardiovascular diseases, or use of therapeutics which may influence circulating blood levels of AGEs or sRAGE derivates.

In our study conducted among the general Western European population, we did not find a clear association of elevated CML or the endogenous secretory receptor esRAGE with risk of pancreatic cancer. Because only 2 prospective studies investigated this relationship but with different soluble RAGE derivates, and one of them exclusively among male smokers, further studies are needed to evaluate the potential role of the AGE/RAGE axis with pancreatic cancer risk.

No potential conflicts of interests were disclosed.

The authors thank Dr. Susen Becker and Britta Lederer for their excellent work in managing and conducting the immunoassays.

V.A. Gorte was funded by a grant from the German Research Foundation, Graduiertenkolleg 793: Epidemiology of communicable and chronic noncommunicable diseases and their interrelationships. This work was supported by WCRF UK and WCRF International, grant no. 2009/39. The coordination of EPIC is financially supported by the European Commission (DG-SANCO) and the International Agency for Research on Cancer. The national cohorts are supported by Danish Cancer Society (Denmark); Ligue Contre le Cancer, Institut Gustave Roussy, Mutuelle Générale de l'Education Nationale, Institut National de la Santé et de la Recherche Médicale (INSERM; France); Deutsche Krebshilfe, Deutsches Krebsforschungszentrum (DKFZ) and Federal Ministry of Education and Research (Germany); Hellenic Health Foundation and Stavros Niarchos Foundation (Greece); Italian Association for Research on Cancer (AIRC) and National Research Council (Italy); Dutch Ministry of Public Health, Welfare and Sports (VWS), Netherlands Cancer Registry (NKR), LK Research Funds, Dutch Prevention Funds, Dutch ZON (Zorg Onderzoek Nederland), World Cancer Research Fund (WCRF), Statistics Netherlands (The Netherlands); ERC-2009-AdG 232997 and Nordforsk, Nordic Centre of Excellence programme on Food, Nutrition and Health. (Norway); Health Research Fund (FIS), Regional Governments of Andalucía, Asturias, Basque Country, Murcia (no. 6236) and Navarra, ISCIII RETIC (RD06/0020; Spain); Swedish Cancer Society, Swedish Scientific Council and Regional Government of Skåne and Västerbotten (Sweden); and Cancer Research UK and Medical Research Council (United Kingdom).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.