Association between Allergies and Risk of Pancreatic Cancer

This article is featured in Highlights of This Issue, p. 367

Pancreatic cancer is a leading cause of cancer deaths with less than 10% of cases surviving 5 years (1, 2), yet the etiology of pancreatic cancer is not well understood. Cigarette smoking continues to be identified as a strong established risk factor (3–5), and more recently, obesity has been consistently associated with increased risk (6–8). Recent literature suggests that heavy alcohol intake (9, 10), non-O blood type (11), and Helicobacter pylori infection (12) modestly increase pancreatic cancer risk. Although diabetes (13, 14) and pancreatitis (15–17) increase risk, diabetes may also be an early manifestation (13, 18), and pancreatitis is extremely rare. Genetics also plays a role; having a first degree relative with pancreas cancer doubles one's risk (19, 20).

Several epidemiologic studies have examined the association between a history of allergies and pancreas cancer risk (21–29). A meta-analysis of 10 case–control and 4 cohort studies concluded that allergies, especially respiratory allergies (e.g., hay fever), were associated with a statistically significant reduction in pancreatic cancer risk, although no association was observed for asthma (30). A recent review further supports this, and summarized that hay fever is associated with reduced pancreatic cancer risk whereas asthma findings are null, noting that literature on other allergies (food, drug, eczema) is inconclusive (16). A trend in reduced pancreas cancer risk was reported for increased number of allergies (22), but this has not been replicated (28). Some previous studies did not collect detailed allergy information, such as specific type of allergy (23) or age at onset (24, 26, 28). To our knowledge, no previous study collected information on skin prick testing, nor differentiated between allergic and nonallergic asthma. Furthermore, no previous study assessed the use of medications to treat allergy symptoms, yet it is important to rule out medications as a possible explanation of the reported allergy–pancreas cancer risk associations. There is no accepted biologic mechanism about how allergies may reduce the risk of pancreatic cancer, although enhanced cancer immune surveillance has been widely suggested (30–33).

This Canadian population-based case–control study improves upon previous studies to evaluate allergies and pancreatic cancer risk by collecting detailed information on allergy history (age at onset, duration, skin prick test results, subtypes, differentiation between allergic and nonallergic asthma), assessing pancreatic cancer risk factors as potential confounders, and being the first study to assess allergy medications as possible confounders.

Pancreas cancer cases were recruited by the Ontario Pancreas Cancer Study (OPCS)—an ongoing study (since 2003) conducted primarily for genetic research and part of the Pancreatic Cancer Genetic Epidemiology Consortium (34). In 2011, and in collaboration with this study, an expanded epidemiology questionnaire, as well as food and allergy questionnaires, was implemented by the OPCS. This study includes pancreatic cancer cases diagnosed between February 1, 2011 and August 31, 2012; cases were identified through the Ontario Cancer Registry, which uses a rapid-case ascertainment system (electronic pathology reporting; International Classification of Diseases for Oncology Third Edition codes C25.0–25.9, with 25.4, neuroendocrine pancreas excluded). Men and women living in Ontario who were diagnosed with a pathologically confirmed adenocarcinoma of the pancreas or adenocarcinoma metastasis, confirmed as pancreatic cancer by treating physicians, and ages ≤89 years, were approached by the OPCS to participate. Of the 345 participating cases, 252 were diagnosed with a pathologically confirmed adenocarcinoma of the pancreas and 93 had adenocarcinoma metastasis (pathology report), and were subsequently confirmed as pancreatic cancer cases by treating physicians.

Pancreas cancer cases were mailed a study package following passive physician consent. This package included 4 self-administered questionnaires: epidemiology, family history, allergy, and food (short 55-item). Cases were also asked if they would be willing to provide a blood sample. Within 2 weeks of the questionnaire mailing, a follow-up postcard was sent to cases who had not returned the package. Telephone calls were made 2 weeks later to nonresponders. Telephone follow-up continued for several months in an effort to improve response rates. If no response was received after 10 weeks, a second questionnaire package was mailed. For cases that became deceased during this follow-up period, a proxy was requested to complete the questionnaires (n = 37 proxies).

A total of 1,095 cases of pancreatic cancer were diagnosed between February 1, 2011 and August 31, 2012 and of these, 327 (30%) were deceased or ineligible (e.g., because of language barrier) leaving 768 eligible cases. The study was not able to mail questionnaire packages to 130 of these cases (57 had no address available, 46 had physicians that refused consent, and 27 cases had no known physician to obtain consent). Thus, 638 cases were mailed the study package, and of these, 137 (21%) refused, 30 (5%) provided only DNA, 126 (20%) were nonresponders, and 345 cases completed study questionnaires, resulting in a 54% response rate. The median lag between pancreas cancer diagnosis date and questionnaire completion was 3 months.

Population-based controls were identified using a modified random digit dialing procedure of households in Ontario and frequency matched (1:3) within 5-year age/sex groups to the expected case distribution. Throughout 2011, control recruitment was conducted by the survey research unit at the Institute for Social Research at York University (Toronto, Ontario). A sampling frame of phone numbers was derived from telephone directories and commercially available lists. Numbers that fell between or on either side of listed numbers were added to the sampling frame to capture unlisted/new numbers. In total, 11,629 households were telephoned; 5,054 households did not have an eligible member, 484 did not answer the phone, 31 did not provide age, and 4,065 refused (eligibility of these households unknown). In total, 1,995 households with an eligible man or woman in a required age-group were identified (one person from each household was randomly selected). Of the 1,995 eligible participants, 261 (13%) persons refused and 1,734 persons agreed to participate in our Ontario Cancer Risk Factor Study (87% response rate). These 1,734 persons (controls) were then mailed a study package that included 3 self-administered questionnaires: epidemiology, allergy, and food (short 55-item). Controls were also sent a form asking whether they would be willing to donate a blood/saliva sample. A reminder postcard was mailed 2 weeks after the questionnaire package. Telephone follow-up of nonresponders was initiated after 4 weeks, and again after 8 weeks. If no response was received after 10 weeks, a second questionnaire package was mailed. Questionnaires were completed by 1,285 controls, for a response rate of 74%. Research Ethics Board approval for this study was obtained from the University of Toronto and Mount Sinai Hospital, Toronto, Canada.

The allergy questionnaire collected information about lifetime history of allergies, including hay fever, eczema/atopic dermatitis, and allergies to dust/mold, animals/pets, insect bites/stings, food, medication, chemicals/metals. Subjects that reported having any of these allergies were asked more detailed questions about age at onset, duration, self-reported results of skin prick testing, and use of medications to treat allergies (oral, eye drop, nasal spray, inhaled; antihistamine, steroid, leukotriene modifier, mast cell stabilizer, β2 adrenergic agonists). The allergy questionnaire also asked about lifetime history of asthma, including type (allergic or nonallergic) and medications used to treat asthma. In addition, all participants were asked about use of antihistamines, nasal sprays, or steroids for reasons other than allergies/asthma. Eight pages of allergy/asthma medication photos were included in the allergy questionnaire appendix to help subjects with recall.

Subjects were categorized as yes or no to ever having had each of the allergies queried; those reporting a particular type of allergy were further categorized based on their skin prick testing results (whether they tested positive to an allergen belonging to that type of allergy). “Any allergy” was defined as having reported any of the 8 types of allergy (asthma not included); participants who responded no to all allergy types were categorized as not having any allergy. A composite variable reflecting respiratory allergies/allergic rhinitis was defined as having hay fever, dust/mold, or animal/pet allergy (35), because respiratory allergy was most strongly associated with reduced pancreas cancer risk in previous studies (30). “Atopy” was defined as a history of hay fever, dust/mold allergy, animal/pet allergy, eczema/atopic dermatitis, or allergic asthma because these conditions are likely the most valid indicators of immunoglobulin E (IgE)–mediated allergy; other allergies (chemical/metal, medication, food, insect) were not included as “atopy” because some persons reporting these conditions likely had intolerances (e.g., to foods or medications) or type IV hypersensitivity reactions (e.g., to nickel), rather than IgE-mediated allergies (misclassified). Allergy history was also categorized based on onset and duration, and on the number of allergic conditions reported. Medications used for the treatment of allergies or asthma were classified according to type (antihistamine, corticosteroid, leukotriene modifier, mast cell stabilizer, β2-adrenergic agonist) and route of administration (oral, inhaled, nasal spray, eye drop).

The distributions of subject characteristics, established pancreas cancer risk factors, and allergy history were described among pancreatic cancer cases and population controls. Multivariate logistic regression was used to estimate the age-adjusted odds ratios (AOR) and 95% confidence intervals (CI) for the association between allergic diseases and pancreas cancer risk; conducted using SAS version 9.2 (SAS Institute Inc.). Possible confounding by pancreas cancer risk factors [family history of pancreas cancer, body mass index (BMI), pancreatitis, type 2 diabetes, blood type, cigarette smoking, alcohol consumption] was evaluated by adding each potential confounder to the regression models and assessing the impact on the AORs for any allergy, respiratory allergy (hay fever, dust/mold, or animal/pet allergy), and atopy. A variable was considered a confounder if any of the 3 AORs changed by >10% (36); none of the variables met this definition and therefore the AOR is presented in all tables. For completeness, Table 2 also contains the multivariate odds ratio (MVOR) in which BMI, diabetes, and smoking variables were forced into the AOR model. Confounding by medications used to treat allergies/asthma (antihistamines, corticosteroids, leukotriene modifiers, mast cell stabilizers, β2-adrenergic agonists) was similarly assessed. Some medications were minimal negative confounders of the allergy associations (changed AOR away from the null); however, we elected to be conservative and not include these medications as covariates so as not to overestimate the strength of association observed between allergies and pancreas cancer risk. Smoking status was assessed as an effect modifier of the allergy–pancreas cancer relationship.

The distributions of subject characteristics and established pancreas cancer risk factors among cases and controls are shown in Table 1, along with AOR estimates. As expected, family history of pancreas cancer, pancreatitis, type 2 diabetes, non-O blood type, and cigarette smoking were associated with increased risk of pancreas cancer.

Table 2 contains the individual and combined allergy variables. Hay fever was associated with a significant reduction in pancreatic cancer risk (AOR = 0.68; 95% CI, 0.52–0.89), and this reduction was even greater for those whose allergy skin prick test was positive for hay fever allergens (AOR = 0.43; 95% CI, 0.26–0.72). Similarly, having a positive dust/mold allergy skin prick test was associated with a statistically significant risk reduction (AOR = 0.49; 95% CI, 0.31–0.78); however, self-reported dust/mold allergy was of borderline significance. Animal/pet allergy was associated with a significant reduction in pancreatic cancer risk (AOR = 0.68; 95% CI, 0.46–0.99), and this reduction was even greater for those who were skin prick test positive. Hay fever, dust/mold, and animal/pet allergies were reported by 37%, 21%, and 17% of the controls, respectively. No statistically significant associations were found between pancreatic cancer risk and each of the following allergies: insects, food, medication, chemical/metal, and eczema/atopic dermatitis. In addition, asthma was not associated with pancreatic cancer risk. All combined/derived allergy variables were associated with reduced pancreatic cancer risk. For example, hay fever, dust/mold, or animal/pet allergy (“respiratory allergy”) was significantly associated with reduced pancreatic cancer risk (AOR = 0.66; 95% CI, 0.51–0.86).

Table 3 contains age of allergy onset, time since allergy onset, and allergy duration for several individual and combined allergy variables. For hay fever, no particular patterns were observed as age at onset, time since onset, and duration all showed consistent associations with reduced risk. No particular pattern was observed as regards dust/mold allergy or animal/pet allergy age at onset, time since onset, and duration, with all point estimates below 1, although as expected with reduced power CIs were wider. Of note, the magnitude of the association was strongest for age at onset 18 to 35 years for both animal allergy and dust allergy. No statistically significant associations were seen with eczema/atopic dermatitis, although most point estimates were below 1. For the combined hay fever, dust/mold, or animal/pet allergy variable, a slightly stronger reduction in risk was observed as age at allergy increased, with an OR of 0.68 for age onset less than 18 years of age, and 0.51 for ages 36 or older.

Having more than one allergic condition conferred a similar reduction in pancreatic cancer risk as having only one allergic condition (Table 4). For example, having one atopic condition is associated with reduced pancreatic cancer risk (AOR = 0.67; 95% CI, 0.49–0.92), as is having 3 or more atopic conditions (AOR = 0.62; 95% CI, 0.41–0.94).

Smoking status showed no statistically significant interaction (P > 0.05) with individual nor combined allergy variables and pancreatic cancer risk (data not shown); however, after stratification, the reductions in risk associated with allergies were more evident among ever smokers as compared with never smokers [e.g., hay fever allergy: ever smoker (AOR = 0.57; 95% CI, 0.39–0.82) versus never smoker (AOR = 0.91; 95% CI, 0.60–1.37)].

Our current finding that self-reported hay fever was associated with a significant reduction in pancreatic cancer risk is consistent with the growing body of literature on this topic (21, 22, 28, 29). We reported a halving of pancreatic cancer risk for persons having a positive skin prick test for hay fever allergens; we are the first to assess this. Similar to other studies, we found that animal/pet allergies (22, 28, 29) and dust/mold allergies (22) were also associated with a significant reduction in pancreatic cancer risk. We did not observe a statistically significant association between eczema/atopic dermatitis and pancreas cancer risk, which is consistent with most previous studies (21, 27, 29), although others have reported a reduction in risk (22, 24). Consistent with prior studies, we did not observe a reduction in pancreas cancer risk associated with the less common allergies to medications (21, 25, 28, 29), insect stings (21, 22), and foods (22, 25, 28, 29), although it is possible conditions were misreported as allergies when they were really intolerances or other non-IgE–mediated reactions. Similar to one study (28), but not another (22), we found that having more than one allergic condition conferred a similar reduction in pancreatic cancer risk as having only one allergic condition. Finally, similar to the 2 other studies to assess interaction with smoking status (29, 37), we found no statistically significant interaction between smoking, allergies, and pancreatic cancer risk, although the reductions in risk were more evident among ever smokers as compared with never smokers. Because we did not observe a pattern with duration of allergy our findings should be interpreted cautiously and need to be replicated.

Previous studies to examine the association between asthma and pancreas cancer risk had slightly inconsistent findings, with a meta-analysis reporting no association (30). Our findings are consistent with the majority of studies reporting no association (e.g., 21, 23, 27, 29), although 2 studies found a reduced pancreas cancer risk associated with asthma (22, 24). The majority of asthma is allergic (triggered by allergens; skin prick test positive), however, a subset of asthma is nonallergic (e.g., triggered by cold air, exercise) and typically has a later onset (38). Previous studies did not differentiate between these 2 types of asthma, thus it is possible that if only allergic asthma is associated with cancer risk and all asthma was captured then misclassification may be responsible for the inconsistent findings. Our study is the first cancer study to differentiate between allergic asthma and nonallergic asthma. Although our study collected information on the type of asthma, we still did not observe an association with pancreas cancer risk, suggesting asthma may not be associated with pancreatic cancer risk in the same way atopic conditions such as hay fever and allergies to pets/dust/mold seem to be.

Allergies are hypersensitivity disorders of the immune system occurring in response to allergens (e.g., pollen), and include hay fever, allergic rhinitis, atopic dermatitis, and atopic asthma. Allergy is mediated by IgE antibodies; allergen-specific IgE is produced by B cells in response to the release of cytokines (e.g., interleukins) by type 2 helper T cells (Th2) that have encountered the allergen (39, 40). IgE antibodies bind to receptors on mast cells and other effector cells, and subsequent exposure to the allergen causes the release of chemicals, such as histamine, which results in allergic inflammation (39, 40). Although it is not well understood why allergies develop, both genetic predisposition (41, 42) and environmental influences are important (43). The hygiene hypothesis suggested several decades ago proposes that childhood infections protect against the development of allergy, and thus the development of allergic disease may be a consequence of reduced microbial exposure, because of improved sanitation, vaccination, and antimicrobial agents (44, 45). A lack of Th1-activating exposures to microorganisms may skew the helper T cell (Th) immune response toward Th2-driven allergy (46). In addition, climate, housing, and lifestyle may alter exposure to allergens, and thus risk of allergic disease (47–49), and pollutants may interact with allergens to enhance IgE-mediated immune responses (50, 51).

There is no accepted biologic mechanism about how allergies may reduce the risk of pancreatic cancer, although it has been suggested that persons with allergies may display enhanced tumor immune surveillance (30–33), where the immune system identifies (via tumor-specific antigens or release of stress molecules) and then eliminates cancerous cells (52, 53). Although there has been some speculation about possible biologic mechanisms, none are established. Allergic conditions have generally been associated with higher levels of allergen-specific IgE (54), and it has been suggested that IgE antibodies may recognize tumor antigens because the molecular epitope pattern determines whether an allergen elicits IgE synthesis/binding, and tumor cell antigens may exhibit similar epitope patterns, and therefore may also trigger IgE responses (55). IgE antibodies have been shown to be elevated in pancreatic cancer cases, compared with controls, and when isolated from patients with pancreatic cancer have been shown to target pancreatic cancer antigens with subsequent cytotoxicity against pancreatic cancer cells (56). It has also been suggested that cells with IgE receptors (e.g., eosinophils, mast cells, dendritic cells) may act as antibody-dependent effector cells in an antitumor immune response (57, 58); these cells are involved in allergic inflammation (59), and are also recruited to the tumor microenvironment (60). Stimulated eosinophils release cytotoxins that can kill tumor cells, and it has been proposed that eosinophils from allergic individuals may be more effective at inducing tumor cell death than those from nonallergic individuals (61, 62). Mast cell involvement with the allergy–cancer relationship is unclear as they release chemicals that may induce tumor cell death [e.g., TNF and interleukin (IL)-4] but can also release VEGF that promotes growth (63, 64). Dendritic cells present tumor antigens to, and activate cytotoxic lymphocytes [e.g., cytotoxic T (T_C) cells, natural killer (NK) cells] that are central to the innate immune system in eliminating tumor cells (53, 65). Dendritic cells express IgE receptors, and it has been hypothesized that IgE could mediate the cross-presentation of tumor antigens by dendritic cells, and thus elicit strong T_C-cell response (66). Although the role of NK cells in allergy is not well understood (67), one recent study found individuals with allergic rhinitis exhibit a higher proportion of NK cells, and greater NK cytotoxicity, than nonatopic controls (68); thus, one could speculate that allergies may be related to the enhanced ability of innate immune system cells to eliminate tumor cells. Finally, it has been suggested that allergy symptoms such as sneezing may serve to expel particles and adhering carcinogens (69, 70).

There is an inherited component to allergies (41, 42), and future studies that comprehensively assess the possible link between variants in immunologic genes and pancreatic cancer risk are warranted. It is interesting that a recent pathway-based analysis of pooled GWAS data reported that Th1/Th2 immune response genes (in particular variants in TGFBR2, CCL18, IL13RA2) were associated with pancreatic cancer risk (71). To our knowledge, the only other assessment of immunologic variants was a small study that found no association between 3 IL-4 gene variants and pancreatic cancer risk (28).

A limitation of prior studies was the inability to evaluate allergy medications as possible confounders. Many prescription and over the counter medications are used to reduce allergy symptoms (antihistamines, corticosteroids, leukotriene modifiers, mast cell stabilizers; ref. 72). It is possible that the protective effect observed between allergies and pancreas cancer risk may be because of medications taken to relieve allergy symptoms; however, this has never been assessed. It has been reported that cromolyn (one type of mast cell stabilizer) reduced pancreatic tumor growth in mice (73), although it is rarely prescribed. It is important to rule out allergy medications as a possible explanation (confounder) of the allergy–pancreas cancer risk association. Our study was the first to collect detailed information on the use of each of these medications and assess confounding by allergy medication use. We found that the reduction in pancreatic cancer risk observed for persons with allergies was not because of allergy medications. Another important strength of our study is that information on skin prick testing was obtained, and this has never been assessed in prior pancreatic cancer studies, although it is likely a more reliable measure of allergy, and as expected the magnitude of associations observed in our study were strongest for those with skin prick positive results reported.

There are a few limitations to this study. Although it is possible self-reported allergic conditions may result in misclassification of atopic status as defined by serum IgE levels (74), self-reported lifetime prevalence is the best measure in the context of a case–control study, as postdiagnostic biologic measures are not relevant to risk. Cohort studies with prediagnostic IgE or skin prick test measures would be ideal; however, IgE/skin prick tests are not routinely conducted in cancer cohort studies. Self-reported hay fever has been shown to be highly correlated with sensitization to plant pollens (75), suggesting self-report may be a valid measure of allergy; however, no formal validation study has been published. Selection bias is always a concern in case–control studies; however, the distribution of all but one risk factor was as expected suggesting our dataset was unbiased. The one exception was BMI (1 year ago), and this may be because a large proportion of our subjects were elderly (≥65 years), and BMI is not an ideal measure of obesity because of loss of muscle mass and changes in the distribution of body fat (76). It is also possible that case–control studies may underestimate the association with BMI because of the weight loss patients with pancreas cancer undergo before diagnosis (77). The prevalence of allergies among our controls is consistent with other North American estimates, suggesting our sample was not biased; the lifetime prevalence of any allergic condition (hay fever, rhinitis, allergy, eczema) in the US NHANES survey was 53% (74), which is comparable to our observed prevalence of 53% for any atopy.

These findings support the growing body of evidence suggesting certain allergies are associated with reduced pancreatic cancer risk. Future research directions that may help elucidate the possible association between allergies and pancreatic cancer include the conduct of cohort studies that evaluate biologic measures of allergy (e.g., serum IgE) and subsequent cancer risk. In addition, while smoking status showed no significant interaction with allergies and pancreatic cancer risk, the risk reductions were more evident among ever smokers as compared with never smokers and future studies with larger sample sizes and greater power should assess this further. Finally, further investigation of the possible link between variants in immunologic genes and pancreatic cancer risk is reasonable.

No potential conflicts of interest were disclosed.

Conception and design: M. Cotterchio, T.J. Hudson

Development of methodology: M. Cotterchio, S. Gallinger

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Cotterchio, E. Lowcock, S. Gallinger

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Cotterchio, E. Lowcock, T.J. Hudson, C. Greenwood

Writing, review, and/or revision of the manuscript: M. Cotterchio, E. Lowcock, T.J. Hudson, C. Greenwood, S. Gallinger

Study supervision: M. Cotterchio

The authors thank A. Borgida for her dedication, as project coordinator, to the Ontario Pancreas Cancer Study.

This research was supported by the Canadian Institutes of Health Research (grantMOP-106631 to M. Cotterchio), and the National Institutes of Health (RO1 CA97075 to S. Gallinger, as part of PACGENE consortium). T. Hudson is a recipient of a Senior Investigator Award from the Ontario Institute for Cancer Research, through generous support from the Ontario Ministry of Research and Innovation.

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.