Inflammatory Bowel Disease and Pancreatic Cancer

A Scandinavian Register-based Cohort Study 1969-2017

Åsa H. Everhov; Rune Erichsen; Michael C. Sachs; Lars Pedersen; Jonas Halfvarson; Johan Askling; Anders Ekbom; Jonas F. Ludvigsson; Henrik Toft Sørensen; Ola Olén


Aliment Pharmacol Ther. 2020;52(1):143-154. 

In This Article


We found patients with IBD to be at increased risk of pancreatic cancer and pancreatic cancer death compared to matched reference individuals. The incidence rate for pancreatic cancer was 22.1 in the patients vs 16.6 in the population, and the HR was 1.43, with the highest excess risk (HR 7.55) in IBD patients with PSC. Once diagnosed with pancreatic cancer, IBD patients were not at increased risk of dying earlier than reference individuals also diagnosed with pancreatic cancer.

A previous meta-analysis found a non-significantly decreased risk of pancreatic cancer with a pooled standardised incidence ratio of 0.51 (0.06–4.57) for Crohn's disease and 0.75 (0.30–1.87) for ulcerative colitis.[24] The six previous population-based cohort studies identified a total of 26 pancreatic cancers in 29 215 patients between 1958 and 2014[15–17,23,33,34] (Table S1), whereas our investigation was based on 442 pancreatic cancers in 161 926 patients between 1964 and 2017.

A Korean study compared cancer cases in patients with IBD diagnosed 2011–2014 with the incidence rate of cancer in the general population in 2013.[17] The pancreatic cancer risk was found to be increased only in women with Crohn's disease (standardised incidence ratio 8.58 [1.04–31.00]), but this analysis was based on only two cases. When stratifying for sex and IBD subtype in our data, there was a statistically significant risk increase for both men and women in Crohn's disease and IBD-U, but in ulcerative colitis only for men.

Patients with PSC have been found to be at increased risk of pancreatic cancer. In a Swedish study where 79% of the cohort had IBD, patients with PSC had 14 times higher risk of pancreatic cancer than the general population.[31] A study based on the Electronic Medical Record IBD cohort in Boston reported an odds ratio for pancreatic cancer of 11.22 (4.11–30.62) in IBD patients with PSC compared to IBD patients without PSC.[32] In our Scandinavian cohort, the HR for pancreatic cancer in patients with IBD vs the IBD-free population was higher in patients with PSC (7.55) than the overall HR of patients with IBD (1.43) but the highest HR was found in patients recently diagnosed with PSC (duration < 5 years 12.0) and the lowest HR was in patients with duration ≥ 10 years (2.83). It should be noted, though, that >90% of all pancreatic cancer events occurred in IBD patients without a PSC diagnosis. That the risk of pancreatic cancer would decrease with increasing time from PSC diagnosis is biologically implausible and the fact that PSC-associated pancreatic cancer to such a high degree was diagnosed around the time of IBD diagnosis suggests that this finding to some extent is an effect of detection bias.

Another potential source of bias is the fact that patients with PSC have an excess risk of developing intraductal cholangiocarcinoma. Because of their anatomical and histopathological similarity, the distinction between cancer in the head of the pancreas and distal cholangiocarcinoma (accounting for 40% of cholangiocarcinoma cases) is sometimes difficult. A misclassification of cholangiocarcinoma as pancreatic cancer could increase the relative risk of pancreatic cancer in patients with IBD and PSC (and consequently lower the relative risk of cholangiocarcinoma).

Although investigation of the reasons for the increased risk of pancreatic cancer in patients with IBD is beyond the scope of this article, some possible mechanisms deserve mentioning. Pancreatitis is common among patients with IBD[26–30,49] and chronic pancreatitis is associated with pancreatic cancer.[5,6] Medication commonly used in IBD, such as thiopurines and tumour necrosis factor inhibitors, are associated with an increased risk of cancer due to immunosuppression.[50] Chronic, dysregulated and persistent inflammation is associated with increased risks of certain malignancies.[51]

Increased medical surveillance in patients with IBD may lead to earlier detection of cancers, as well as the reverse situation, where cancer symptoms may lead to a diagnosis of IBD. In order to reduce detection bias we also analysed data excluding the first year of follow-up, and found that the incidence rate of pancreatic cancer only decreased marginally, from 22.1 to 20.8 in the patients with IBD and increased slightly from 16.6 to 16.8 in the general population. However, our results otherwise indicate that pancreatic cancer was not detected earlier in patients with IBD than in the population. First, we did not observe a difference in cancer stage between patients and reference population (P = 0.174, Table S9). Second, the proportion undergoing resectional surgery (indicating potentially curable disease) was slightly lower in IBD patients (8.14% vs 9.60% in the reference population). Third, the relative risk of pancreatic cancer death when comparing IBD patients with pancreatic cancer and reference population with pancreatic cancer was not decreased (1.07 [0.95 to 1.21]).

The main strength of this study is its large binational study population based on prospectively recorded data with long follow-up from registers with virtually complete coverage. Due to population-based setting, our results should be highly generalisable to similar populations. Access to histopathology data helped define IBD onset better than in previous reports.[21,39] Unlike some earlier studies using standardised incidence ratios (which might underestimate the relative risk since reference individuals could be diagnosed with cancer before study entry whereas IBD cases could not), our comparisons of cancer incidence were based on a matched reference cohort drawn from the general population.

Study limitations include lack of data on smoking, which is the main risk factor for pancreatic cancer. We therefore used a quantitative bias analysis to assess how sensitive our main results were to unmeasured confounding and found that an unmeasured confounder such as smoking would have to be highly imbalanced between IBD patients and reference individuals to fully attenuate the relationship between IBD and pancreatic cancer risk (Figure S2). The fact that the incidence of pancreatic cancer was similar between Crohn's disease and ulcerative colitis, although the prevalence of past and current smoking differs between IBD subtypes, further corroborates that confounding from smoking unlikely explains our finding of increased risk of pancreatic cancer in patients with IBD. This quantitative bias analysis can be used to assess the importance of any unmeasured confounder (eg family history, alcohol abuse).

Our cohort ranged over a large time span during which access to register data has increased to include also outpatient specialist care. Although the patient cohort was based only on inpatient data during the first period of follow-up (until 1995 in Denmark and until 2001 in Sweden), the variations in incidence were small, in accordance with the fact that pancreatic cancer incidence in the population has been quite constant.[42]