Risk of pancreatitis in patients with inflammatory bowel disease – a meta-analysis

Inflammatory bowel disease (IBD) is an immune-mediated disease characterised by chronic intestinal inflammation with alternating periods of remission and relapse [1-4]. IBD, comprising ulcerative colitis (UC) and Crohn’s disease (CD), is observed predominantly in developed countries; but after many years on the rise, the incidence now seems to have stabilised in the Western countries. Conversely, the incidence is now increasing in newly industrialised countries. Hence, the global burden of IBD is increasing substantially [5, 6].

Pancreatitis is characterised by inflammation in the pancreas and is categorised as either acute or chronic. Both acute and chronic pancreatitis are severe conditions that lead to hospitalisation and possibly death. The overall mortality rate in acute pancreatitis falls in the 2-9% range [7].An association between IBD and pancreatic lesions was initially reported by Ball et al in 1950: in an autopsy study, pancreatic inflammation was observed in 53% of patients with UC compared with 3% in the control group [8]. Subsequently, other studies have reported an association between IBD and pancreatitis [9]; however, a recent case series did not find an increased incidence of pancreatitis in patients with IBD compared with the general population [10].

Several mechanisms have been suggested to underlie a potential association between IBD and pancreatitis [11]. One hypothesis is that pancreatitis is an extra-intestinal manifestation of IBD [12] and thus related to a shared pathogenic pathway. Another hypothesis is that pancreatitis is caused by the management of IBD, especially by medications, or by associated diseases, especially gall stones [9]. The aim of the present study was to investigate the association between IBD and risk of pancreatitis by conducting a systematic literature review and meta-analysis of observational studies reporting on the risk of pancreatitis in patients with IBD as compared with IBD-free individuals.

Methods

A systematic review of the existing literature addressing the association between IBD and the risk of pancreatitis was performed and reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) [13] and the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines [14].

Eligibility criteria

Eligible studies included observational studies investigating a study population of adult patients with IBD and comparing this population with a non-IBD population. Studies were included if the outcome was acute pancreatitis and/or chronic pancreatitis. Studies of children, animal studies, case reports and reviews were excluded.

Information sources and search strategy

To identify relevant studies addressing the risk of pancreatitis among patients with IBD, a literature search with no language restrictions was performed in PubMed and Embase from database inception until October 2018. A variation of synonyms of the exposure ”inflammatory bowel disease”, the outcome ”pancreatitis” and the study design ”epidemiologic studies” was combined into search strings. The complete search strategy including specific search strings is presented in the supplementary figure. Additionally, we systematically reviewed the reference lists of the eligible

studies.

Study selection

Two authors independently screened titles and abstracts. After this initial screening, 14 articles were read in full to identify studies meeting the eligibility criteria.

Data collecting process and data forms

Data extraction was conducted for the studies eligible for inclusion in the systematic review. The data items included were: author, year of publication, country, study design, source population and period of time, sample size, statistical analysis including covariates, and main findings. When both crude and adjusted estimates were presented in the included articles, the adjusted estimates were extracted for the systematic review and meta-analysis.

Risk of bias

Risk of bias was assessed using the Risk of Bias Assessment tool for Non-randomized Studies (RoBANS) [15]. Based on RoBANS, the studies were categorised as having an overall low risk of bias (≤ 1 domain evaluated as having a high or unclear risk of bias), medium risk of bias (two domains evaluated as having a high or unclear risk of bias) or high risk of bias (≥ 3 domains evaluated as having a high or unclear risk of bias). To evaluate how well the studies accounted for potential confounders, a list of major confounding variables was defined. As major risk factors for developing pancreatitis include alcohol use, gall stone disease, and medication, and as these variables may also be associated with IBD, they had to be adjusted for through study design or statistical modelling, for a study to be assessed as “low risk” in the domain “Confounding variables”.

Data synthesis and analysis

The studies included in the meta-analysis reported different types of effect estimates – odds ratios, standardised incidence ratios or hazard ratios – but as these estimates are all closely related to measurements of relative risk and hence support the same interpretation, we found it acceptable to here compare and combine the estimates through meta-analysis. For one study [16], which only reported separate estimates for CD and UC, we derived the estimate for IBD simply by adding the observed and expected number of cases for each subtype, and subsequently calculating the 95% confidence interval (CI) assuming a Poisson distribution of observed cases. Statistical heterogeneity was tested using the chi-squared test and measured by the I² statistic. The effect estimates were combined based on a random effects meta-analysis approach where studies are weighted based on corresponding standard errors [17]. Since the heterogeneity parameter was here consistently estimated to be zero, in practice fixed effects meta-analyses were performed (I² = 0.0%; for all p-values: p ≥ 0.41). Analyses were derived both for IBD and specifically for CD and UC. Furthermore, estimates were combined on the log-risk ratio scale and then finally back-transformed to the original scale. All statistical analyses were carried out in Stata 14.2 (StataCorp LLC, College Station).

Results

Data retrieval

A total of 895 articles were identified through literature search and their titles and abstracts were screened for eligibility (Figure 1). Among those, 881 were excluded as the title and abstract revealed that the studies were not eligible. After full-text reading, an additional ten articles were excluded (Figure 1) and, finally, four studies [16, 18-20] were included in our systematic review and meta-analysis.

Study characteristics

Characteristics of the four included studies are presented in Table 1. All four studies had acute pancreatitis as outcome. The studies were published between 1999 and 2016. Two studies were from Denmark, one from Sweden and the latest study was from Taiwan. Two of the studies were register-based prospective cohort studies. The two other studies were case-control studies, one based on register data and one based on structured interviews. We found no eligible studies with chronic pancreatitis as outcome.

Risk of bias within and across studies

To assess the risk of bias of the included studies, the RoBANS tool was used (Table 2).

All studies were of a high quality and were evaluated as having a low overall risk of bias. The only domain in which all the studies except one were evaluated as having a high risk of bias was the domain regarding confounding variables. The studies included were all relatively new (published in the 1999-2016 period).

The study populations were all based on large study cohorts of patients with a distinct IBD diagnosis and non-IBD subjects (numbers ranging from 2,243 to 59,545), based on register data or from structured interviews. The age and sex distribution of the study population was comparable across the four studies.

All studies were from developed countries with three studies based on Scandinavian study populations and one study based on an Asian study population. Funnel plots did not indicate asymmetry, but plots based on few studies should be interpreted with caution.

Risk of acute pancreatitis in patients with inflammatory bowel disease

Two studies provided risk estimates of acute pancreatitis for IBD patients overall [18, 20]. Additionally, for one study [16], which only reported separate estimates for CD and UC, we were able to derive the estimate for IBD overall as explained in the Methods section. Risk estimates of acute pancreatitis for IBD patients overall ranged from 2.56 (95% CI: 2.04-3.20) to 3.40 (95% CI: 1.50-7.90). Three studies provided risk estimates of acute pancreatitis for CD and UC [16, 19, 20], with risk estimates ranging from 3.40 (95% CI: 2.70-4.28) to 4.31 (95% CI: 2.86-6.25) for CD and from 1.50 (95% CI: 0.70-3.60) to 2.49 (95% CI: 1.91-3.26) for UC (Table 1).

The studies included in this systematic review all found an increased risk of acute pancreatitis in patients with IBD. Accordingly, the overall estimated risk ratio revealed a significantly increased risk of acute pancreatitis in patients with IBD of 2.78 (95% CI: 2.40-3.22; p < 0.001) (Figure 2). The risk of acute pancreatitis was significantly increased for both CD and UC, with the highest risk observed in patients with CD: the overall estimated risk ratio was 3.62 (95% CI: 2.99-4.38; p < 0.001) for patients with CD and 2.24 (95% CI: 1.85-2.71; p < 0.001) for patients with UC.

Discussion

The present systematic review and meta-analysis shows that patients with IBD are at increased risk of developing acute pancreatitis compared with non-IBD individuals. The risk is most pronounced in patients with CD, who are at a three-fold increased risk of developing acute pancreatitis, whereas patients with UC have a two-fold increased risk. We identified no eligible studies with chronic pancreatitis as outcome.

To our knowledge, this is the first systematic review and meta-analysis investigating the risk of pancreatitis in patients with IBD compared with non-IBD individuals. Previous reviews have included case series or focused on the spectrum of pancreatic disorders in patients with IBD [9, 11]. Another strength of the present study is our thorough literature search with two authors screening all titles and abstracts. Furthermore, the included studies were systematically evaluated to ensure that only studies of a high quality were included in the meta-analysis. Using the RoBANS tool for bias

assessment, we found all included studies to have an overall low risk of bias. However, three of the four studies were evaluated as having a high risk of bias in the specific domain of confounding variables, wherefore lack of control for confounding represents a potential limitation to this systematic review. Moreover, only four studies met the inclusion criteria, limiting evaluation of publication bias and subgroup effects.

The mechanisms underlying the increased risk of acute pancreatitis, especially in patients with CD, are not well described. Pancreatitis may represent an extra-intestinal manifestation of IBD [12] related to a shared pathogenic pathway. This type of pancreatitis comprises idiopathic pancreatitis but also autoimmune pancreatitis, which has been associated with IBD in recent years [21, 22]. Medical treatment of IBD, especially

azathioprine [23], or co-morbidities per se, particularly gallstones, have been shown to lead to an increased risk of pancreatitis [9]. However, two of the studies [16, 20] included in our meta-analysis did not take information on medication use into account, and two studies [18, 19] included only information on selected medication use.

Therefore, the medications used for treatment of IBD likely contribute to the increased risk of pancreatitis observed in our meta-analysis. In support of this notion, a recently published Swedish-Danish nationwide cohort study of children with IBD showed a considerably increased risk of acute pancreatitis among children initiating azathiopurine therapy [24]. The increased risk of pancreatitis found in patients with IBD could also be explained by surveillance bias, as patients are followed closely and examined thoroughly in the course leading to IBD diagnosis and during the period after IBD diagnosis. A diagnosis of acute pancreatitis is generally based on two of the following three criteria [25]: 1) acute and sudden abdominal pain; 2) increase in serum amylase and/or lipase greater than > 3 times upper limit; 3) characteristic findings on contrast enhanced-computed tomography/ultrasound/magnetic resonance.

However, if surveillance bias was to explain the observed association between IBD and pancreatitis, one would expect the risk to be similar among patients with CD and UC, as there are no major differences in examination and follow-up programmes for the two subtypes of IBD. Moreover, as one would expect patients with clinical symptoms of pancreatitis to seek medical help due to pain, regardless of an IBD diagnosis, surveillance bias seems less likely.

CONCLUSIONS

With this systematic review and meta-analysis of observational studies of acute pancreatitis among patients with IBD as compared with non-IBD individuals, we conclude that patients with IBD are at increased risk of developing acute pancreatitis. The risk of acute pancreatitis is three-fold increased in CD and two-fold increased in UC, whereas the risk of chronic pancreatitis in patients with IBD remains unknown. The studies

included in the present meta-analysis only provide limited information on factors possibly contributing to the observed increased risk of pancreatitis in IBD. Hence, it is of clinical relevance to further study the impact of disease severity, co-morbidities, surgery and IBD medications including biologic drugs on risk of acute pancreatitis among patients with IBD. Moreover, studies of the risk of chronic pancreatitis among patients with IBD are warranted.

Correspondence: Kristine H. Allin. E-mail: kristine.allin@regionh.dk

Accepted: 24 January 2020

Conflicts of interest: none. Disclosure forms provided by the authors are available with the full text of this article at Ugeskriftet.dk/dmj