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


Characteristics of the Study Population

We identified 161 926 patients with incident IBD and no previous diagnosis of pancreatic cancer from Denmark and Sweden, who were matched to 1 599 024 IBD-free reference individuals from the general population (Figure 1). The majority (70%) of patients were from Sweden. Ulcerative colitis was more common (60%) than Crohn's disease (29%) and IBD-U (11%) (Table 1). Median age at IBD diagnosis was 38 years, 51% were followed for 10 years or more and 2.1% were diagnosed with PSC at or during follow-up (Table S6). Descriptive statistics by country are found in Table S6.

Figure 1.

Flow chart of inclusion and exclusion criteria

Pancreatic Cancer

During a follow-up of 2 000 951 person years, 442 patients with IBD (0.27%) were diagnosed with pancreatic cancer compared to 3386 (0.21%) individuals from the reference population during 20 402 533 person years (Table 2). The 10-year cumulative incidence was 0.18% (95% CI: 0.16–0.20) in patients and 0.14% (0.13–0.15) in the reference population (difference 0.04% [0.02–0.06]), and the 20-year cumulative incidence was 0.34% (0.30–0.38) vs 0.29% (0.28–0.30) (difference 0.05% [0.01–0.09]). The incidence rate was 22.1 (20.1–24.2) events/100 000 person years in the patients and 16.6 (16.0–17.2) in the reference population. The standardised incidence rate was 9.04 (8.15–9.93) in patients vs 7.02 (6.77–7.26) in the matched reference individuals and more common for both patients with IBD and matched reference population in Denmark than in Sweden (Table S7).

The adjusted HR for pancreatic cancer in patients with IBD compared to the reference population was 1.43 (1.30–1.58) and decreased to 1.34 (1.20–1.48) excluding the first year of follow-up (Table 3). The HR was 1.44 (1.18–1.74) in patients with Crohn's disease, 1.35 (1.19–1.53) for ulcerative colitis and 1.99 (1.50–2.64) for IBD-U. The HR was 1.56 (1.37–1.79) in men and 1.30 (1.12–1.50) in women. The HR point estimates were highest in patients diagnosed with IBD as children: 2.78 (1.13–6.86) and lowest in those diagnosed as elderly (≥60 years): 1.33 (1.14–1.56), but with overlapping 95% CIs. IBD patients with PSC were at increased risk of pancreatic cancer with 34/442 (8%) of pancreatic cancer cases in IBD occurring in IBD patients with a PSC diagnosis. The overall HR in IBD patients with PSC was 7.55 (4.94–11.5) and the risk decreased with the duration of concurrent PSC (HR duration <5 years: 12.0 [6.42–22.3], duration 5–10 years: 10.1 [4.06–25.3], duration ≥10 years: 2.83 [1.09–7.32]). Stratified risk estimates by IBD subtype are presented in Table S8.

Pancreatic Cancer Stage, Surgery and Death

Cancer stage was not different between patients with IBD and the reference population (P = 0.17), and less than 10% of both IBD patients and reference individuals underwent resectional surgery of the pancreas (Table S9). In the multi-state model, patients with IBD were at increased risk of pancreatic cancer, pancreatic cancer death and death from other causes. However, IBD patients with a diagnosis of pancreatic cancer were not at increased risk of dying from pancreatic cancer compared to reference individuals also diagnosed with pancreatic cancer (Figure 2, Figure 3, Table S10). The results were similar when stratifying for IBD subtype (Crohn's disease: Figure S1 and Table S11; ulcerative colitis: Figure S2 and Table S12). The cumulative additive effect of IBD on the hazard for pancreatic cancer diagnosis in patients with IBD vs reference individuals increased with time from diagnosis, as did the risk of dying from other causes (Figure 4). The relative pancreatic cancer-free survival rate decreased with time from diagnosis, but we did not observe any dramatic changes across calendar periods (Figure 5). The HR for pancreatic cancer diagnosis restricted to all study participants at risk during the last 10 years of follow-up was 1.34 (1.01–1.77) for Crohn's disease, 1.30 (1.09–1.55) for ulcerative colitis and 1.64 (1.04–2.59) for IBD-U.

Figure 2.

Multi-state model showing the states under consideration and the possible transitions between them. The thickness of the transition lines is proportional to the log of the number of transitions

Figure 3.

Cause-specific cumulative incidence curves for each of the transitions in the multi-state model

Figure 4.

Adjusted time varying effect of inflammatory bowel disease on transition hazards

Figure 5.

Time-trends of hazards for pancreatic cancer comparing patients with incident inflammatory bowel disease with matched general population reference individuals by follow-up time and calendar year at diagnosis

Quantitative Bias Analysis

Smoking has been positively associated with pancreatic cancer with an HR of 1.74 (95% CI 1.61–1.87) for current smokers and 1.20 (95% CI 1.11–1.29) for former smokers.[3] The prevalence of daily smoking in Sweden was 14% in 2006 and 7% in 2018[47] and daily smoking in Danish adults (>15 years old) was 16% in 2015.[48] An unmeasured binary confounding variable (eg smoking status) with an HR for pancreatic cancer diagnosis of two (ie even higher than described) would need to have at least a 60% difference in prevalence between patients with IBD and the reference population, that is, 80% among IBD and 20% in the reference population, to lower the estimated HR for IBD to 1.00 (Figure S2). It is, therefore, extremely implausible that smoking explains our main findings more than marginally.