The incidence of pancreatic cancer (PC) has been increasing in the US and worldwide.^{1, 2} Pancreatic ductal adenocarcinoma (PDAC) and pancreatic neuroendocrine tumors (PNETs) are the two most common PC subtypes. Cigarette smoking, diabetes mellitus, chronic pancreatitis and excess weight are known risk factors for PDAC.³ Risk factors for PNETs are largely unknown beyond rare inherited genetic syndromes.⁴

Autoimmune conditions are characterized by immune dysregulation and have been associated with increased risks of gastrointestinal tract cancers in registry-based studies.^5–18 However, current evidence on the associations of autoimmune conditions, particularly Crohn’s disease,^{6, 7, 10, 16, 17} ulcerative colitis,^{7, 8, 10, 16, 17, 19} and systemic lupus erythematosus,^{10, 15} and celiac disease^{10–12, 16} have been inconsistent. Many of these studies have insufficient power to detect associations due to the relatively low prevalence of autoimmune conditions and PC in the general population. Previously, we showed that joint effects of common variants in genomic regions containing susceptibility loci for Crohn’s disease and ulcerative colitis were associated with PDAC, suggesting a shared genetic architecture between the diseases.²⁰

In this study, we examined associations between autoimmune conditions and PC risk in a large population-based, nested case-control study of older adults within the US Surveillance, Epidemiology, and End Results Program (SEER)-Medicare linked data. We hypothesize that autoimmune conditions are associated with the risk of PC.

Our study included 80,074 incident PC cases and 320,296 population-based controls (Table 1). Among the PC cases, 61,081 (76.3%) were first primary and 18,993 (23.7%) were later primary. Compared with controls, PC cases were more likely to be non-Hispanic Black, have longer Medicare Part A, B, non-HMO coverage, visit physicians more frequently six months prior to diagnosis/selection, and have diagnoses of overweight/obesity, COPD, acute/chronic pancreatitis, type 2 diabetes, and smoking behavior- and alcohol-related comorbidities. Compared with first primary cases, later primary cases were more likely to be older, male, non-Hispanic White, have longer Medicare Part A, B, non-HMO coverage and more physician visits six months prior to diagnosis, and have diagnoses of overweight/obesity, COPD, acute/chronic pancreatitis, type 2 diabetes, and smoking behavior- or alcohol-related comorbidities.

Eight autoimmune conditions were significantly associated with an elevated risk of all PC at P_FDR value < 0.10 (Table 2): ankylosing spondylitis (OR = 1.45; 95% CI: 1.14–1.84), Graves’ disease (OR = 1.18; 95% CI: 1.03–1.34), localized scleroderma (OR = 1.27; 95% CI: 1.06–1.52), pernicious anemia (OR = 1.08; 95% CI: 1.02–1.14), primary sclerosing cholangitis (OR = 1.37; 95% CI: 1.18–1.59), pure red cell aplasia (OR = 1.31; 95% CI: 1.16–1.47), type 1 diabetes (OR = 1.11; 95% CI: 1.07–1.15), and ulcerative colitis (OR = 1.18; 95% CI: 1.07–1.31). We also observed a suggestive positive association for discoid lupus erythematosus (OR = 1.30; 95% CI: 1.03–1.64; P_FDR value = 0.12). These associations were similar in magnitude or in the same direction for first and later primary PC except for pure red cell aplasia which was only associated with later primary PC (OR = 2.00; 95% CI: 1.68–2.37). In addition, aplastic anemia (OR = 2.09; 95% CI: 1.21–3.61) was positively associated with later primary PC, whereas rheumatoid arthritis (OR = 0.89; 95% CI: 0.81–0.97) was inversely associated. Associations with PC were generally similar in stratified analyses, although stronger associations were present in women compared with men for Graves’ disease; in participants ≥ 75 years old compared with < 75 years old for localized scleroderma, primary sclerosing cholangitis, and pure red cell aplasia; and in non-Hispanic White compared with non-Hispanic Black for pernicious anemia and pure red cell aplasia (online supplementary tables 2–4). In hierarchical models that mutually adjusted for conditions in the same group, the eight top autoimmune conditions remained significantly associated with PC (online supplementary table 5). Among participants diagnosed with autoimmune conditions five years or more before PC diagnosis (online supplementary table 6, n=56,624 cases and 220,531 controls), Graves’ disease (OR =1.33; 95% CI: 1.10–1.62), pure red cell aplasia (OR = 1.45; 95% CI: 1.19–1.77), type 1 diabetes (OR = 1.15; 95% CI: 1.10–1.21), and ulcerative colitis (OR = 1.22; 95% CI: 1.05–1.43) remained significant after adjusting for multiple comparison, whereas the associations for ankylosing spondylitis (OR = 1.61; 95% CI: 1.11–2.35) and localized scleroderma (OR = 1.40; 95% CI: 1.07–1.83) were positive however not at P_FDR value < 0.10. The associations for pernicious anemia (OR = 1.06; 95% CI: 0.97–1.15) and primary sclerosing cholangitis (OR = 0.96; 95% CI: 0.69–1.34) became non-significant.

Table 3 shows the results for the eight most significant, PC-associated autoimmune conditions (P_FDR value < 0.10) stratified by the PDAC and PNET subtypes. Ankylosing spondylitis (OR = 1.42; 95% CI: 1.12–1.81), Graves’ disease (OR = 1.18; 95% CI: 1.04–1.35), localized scleroderma (OR = 1.24; 95% CI: 1.03–1.48), pernicious anemia (OR = 1.08; 95% CI: 1.02–1.14), primary sclerosing cholangitis (OR = 1.38; 95% CI: 1.19–1.60), pure red cell aplasia (OR = 1.33; 95% CI: 1.18–1.49), type 1 diabetes (OR = 1.11; 95% CI: 1.07–1.15), and ulcerative colitis (OR = 1.16; 95% CI: 1.05–1.29) were significantly associated with PDAC. Only ulcerative colitis was associated with PNETs (OR = 1.78; 95% CI: 1.17–2.70).

In this largest, most comprehensive study to date, prior diagnoses of ankylosing spondylitis, Graves’ disease, localized scleroderma, pernicious anemia, primary sclerosing cholangitis, pure red cell aplasia, type 1 diabetes, and ulcerative colitis were significantly associated with 8–45% increased risk of primary PC. The associations were in the same direction for first and later primary PC except for pure red cell aplasia, where a positive association was only present with risk of later primary. In analyses stratified by PC histological subtypes, the eight autoimmune conditions were similarly associated with an elevated PDAC risk, whereas only ulcerative colitis was significantly associated with 78% increased PNETs risk.

The positive associations we observe for ankylosing spondylitis, Graves’ disease, and localized scleroderma with overall PC and PDAC, and for ulcerative colitis with PNETs are novel. Studies based on US Veterans Affairs hospital admissions or registry-based cohort studies in Sweden and Taiwan have not observed significant associations for ankylosing spondylitis, Graves’ disease, or localized scleroderma and PC, although these studies included small numbers of PC cases (N = 1–24).^{10, 27–29} Ankylosing spondylitis has been significantly associated with cancer risk of the ampulla of Vater,¹⁴ which connects pancreatic ducts to the small intestine. Methimazole, an antithyroid medication used to treat Graves’ disease, has been shown to increase the risk of acute pancreatitis and might contribute to the association we observe.³⁰ To the best of our knowledge, no previous study has examined the association of ulcerative colitis with PNETs. One population-based study based on the Dutch Pathology Registry reported a higher prevalence of colonic neuroendocrine tumors among inflammatory bowel disease patients compared with the general population.³¹ Future observational and animal studies are needed to replicate our results and elucidate plausible disease mechanisms.

Previous studies on the association of primary sclerosing cholangitis with PC risk are limited. Two Swedish studies have reported 8 to 14-fold elevated risk for primary sclerosing cholangitis, albeit based on ≤ 20 PC cases.^{9, 18} The observed association may be partially explained by the misdiagnosis of PC with extrahepatic cholangiocarcinoma,³² a malignancy which is strongly associated with primary sclerosing cholangitis.¹⁴ However, this is unlikely a limitation of our study because cancer diagnoses from SEER registries are of the highest data quality³³ and our study included a large number of PC cases. Current evidence on the association of ulcerative colitis with PC risk is inconsistent. Scandinavian registry-based studies have observed significant 35–45% elevated PC risk associated with ulcerative colitis diagnosis.^{7, 8} One case-control study among 1,705 PC cases and 1,084 controls from PanGenEU (European Study into Digestive Illnesses and Genetics) study reported a null association between ulcerative colitis and PC risk,¹⁹ however was based on 15 cases and 16 controls reporting ulcerative colitis. A more recent case-control study in the SEER-Medicare population reported a suggestive positive association between ulcerative colitis and first primary PC risk (OR = 1.12; 95% CI: 0.99–1.27).¹⁷ In contrast to our study, this earlier SEER-Medicare study did not include later primary PC and did not adjust for overweight/obesity, smoking behavior-related diagnoses, alcohol-related diagnoses, type 2 diabetes mellitus, and acute/chronic pancreatitis. Our findings are also consistent with our earlier investigation using genome-wide association study (GWAS) data which suggested that ulcerative colitis shared genetic architecture with PDAC.²⁰ We did not observe a statistically significant association of genomic regions for primary sclerosing cholangitis (P value = 0.078) with PDAC in our earlier analysis, partly due to few established GWAS loci for primary sclerosing cholangitis.²⁰ Patients with inflammatory bowel diseases, particularly ulcerative colitis, are known to have a higher prevalence of primary sclerosing cholangitis,³⁴ and it is possible that the positive PC associations observed for both autoimmune conditions could be driven by the same underlying mechanism of carcinogenesis.

Diabetes is a known risk factor for PC, most notably type 2 diabetes related to excess adiposity.^{35, 36} Growing evidence suggests type 1 diabetes may be associated with PC.^{37, 38} In a meta-analysis based on three cohort and six case-control studies with a total of 39 PC cases, patients diagnosed with type 1 or young-onset diabetes before the age of 40 had a two-fold increased PC risk compared with non-diabetic subjects.³⁷ A population-based registry study from five European countries including 387 PC cases found statistically significant 1.53- and 1.25-fold elevated PC risks among male and female patients with type 1 diabetes, respectively. The associations were attenuated 10 to 15 years after diabetes diagnosis possibly due to reverse causation.³⁸ A limitation of these studies is the potential misclassification of diabetes type because type 1 diabetes was frequently defined as having diabetes at a young age without a specific diagnosis.^{37, 38} Although we used ICD-9 codes to ascertain type 1 diabetes, diabetes misclassification may still be present in our study.

Findings from previous studies on the association between pernicious anemia and PC are conflicting. Two population-based cohort studies and one case-control study reported null associations between pernicious anemia and PC among hospitalized male US veterans, patients in the Health Improvement Network database of the United Kingdom, and participants of the PanGenEU study.^{10, 19, 39} Possible explanations for discrepancies in the reported association between our study and these earlier studies include differences in study design and/or population, sample size, and adjusted covariates in regression models. A Swedish population-based hospital discharge cohort⁴⁰ and an earlier population-based, case-control study in the SEER-Medicare population both found significant positive associations between pernicious anemia diagnosis and PC risk,⁴¹ which are consistent with our findings. Pernicious anemia is an autoimmune condition that results from autoantibody-mediated destruction of intrinsic factor-secreting parietal cells, and manifests primarily as vitamin B₁₂ deficiency, achlorhydria and subsequent elevated serum gastrin concentration.⁴² In vitro studies have shown that gastrin stimulates the proliferation and metastasis of human PC cells.^{43, 44} Given that our association between pernicious anemia and PC became non-significant when we excluded cases with autoimmune conditions first reported less than five years prior to PC diagnosis, our results might also be explained by reverse causation.

The biological basis for our observed associations is speculative. Most of the autoimmune conditions that we observed significant associations with PC affect the gastrointestinal tract either directly or secondarily. Chronic inflammation and subsequent tissue damage caused by autoimmune conditions, or their therapies may promote tumorigenesis and progression through multiple mechanisms, including DNA damage and oncogenic mutations, angiogenesis, antagonizing anti-tumor immunity, cell growth and survival, and enhanced invasion and metastasis.⁴⁵ Immunosuppressive treatments for ulcerative colitis and localized scleroderma,^{46, 47} hyperglycemia and exogenous insulin treatment related to type 1 diabetes,³⁸ and tumor-promoting effect of thyroid hormones for Graves’ disease could also contribute.⁴⁸ The associations of pure red cell aplasia and aplastic anemia with later primary PC risk might be sequelae of the earlier malignancies or cancer therapies, respectively.^{49, 50}

Strengths of our study includes large numbers of PC cases which allowed for examination of rare autoimmune conditions and PC subtype analyses. SEER-Medicare is representative of the US population aged 65 years and older.²¹ The PC cases were identified through SEER registries, which employs rigorous standards to ensure accuracy of cancer case ascertainment.³³ By using Medicare claims for autoimmune conditions identified more than one year prior to cancer diagnosis instead of self-report, we reduced the likelihood of reverse causation and recall bias.

Our study also has limitations. Since many autoimmune conditions causes symptoms that are associated with pancreatic cancer, surveillance bias could potentially affect the study findings (e.g., fatigue and weight loss caused by Graves’ disease and red cell dysplasia could make physicians monitor pancreatic disorders more closely). However, we attempted to minimize surveillance bias by adjusting for the average number of physician visits per 6 months in the 5 years before cancer diagnosis/control selection and excluding autoimmune conditions diagnosed within 12 months of pancreatic cancer diagnosis. Additionally, we conducted a sensitivity analysis in which only autoimmune conditions diagnosed at least 5 years before pancreatic cancer diagnosis were included. Our findings were robust and mostly remained in all these analyses. Given that our study examined the associations of 45 autoimmune conditions with pancreatic cancer risk, some of the significant findings may be due to chance; however, we controlled for multiple comparisons by employing a false discovery rate of < 0.10. Thus, among the 8 findings (or 4 after excluding cancer cases diagnosed within 5 years of autoimmune condition diagnosis), at most 10% could be false positives. Due to the nature of Medicare claims data, some autoimmune conditions (e.g., primary sclerosing cholangitis)⁵¹ and risk factors (e.g., overweight/obesity, smoking and alcohol use) may be under-ascertained, particularly if medical conditions were diagnosed prior to age 66. We applied strict definitions of disease diagnoses (i.e., ≥ one inpatient claim or two physician/outpatient claims) to reduce potential misclassification of autoimmune conditions and would expect under-ascertainment to attenuate associations. We used alcohol- and smoking behavior-related disorders and diseases as proxies for heavy drinker and smokers, and thus some residual confounding may be present. Our study lacks complete information on treatments for autoimmune conditions, which could influence cancer risk.⁴⁶ Finally, our findings may not be generalizable to younger persons with autoimmune conditions.

In conclusion, results from our large, comprehensive analysis support autoimmune conditions, particularly those that affect gastrointestinal tract either directly or secondarily, are associated with increased PC risk in older US adults. The prevalence of autoimmune conditions is increasing in the US, particularly in older people,⁵² and could plausibly be contributing to the increasing incidence of PC in addition to other known PC risk factors. Further population and experimental investigations are warranted to confirm our findings and elucidate the mechanisms by which immune dysregulation contributes to pancreatic carcinogenesis.