Risk Factors for Primary Sclerosing Cholangitis

Kirsten Boonstra; Elisabeth M. G. de Vries; Nan van Geloven; Karel J. van Erpecum; Marcel Spanier; Alexander C. Poen; Carin M. van Nieuwkerk; Ben J. Witteman; Hans A. Tuynman; Anton H. Naber; Paul J. Kingma; Ulrich Beuers; Cyriel Y. Ponsioen


Liver International. 2016;36(1):84-91. 

In This Article


This is the largest case–control study showing that smoking behaviour is independently associated with a lower risk of developing PSC, using a population-based PSC cohort. The negative association between smoking and PSC in this study is in agreement with previous studies, showing odds ratios of having PSC for smokers vs. non-smokers ranging from 0.33 to 0.37.[21–23,27] Interestingly, a recent case–control study by Eaton et al. showed that smoking was negatively associated with PSC when concomitant IBD was present, but not among PSC patients without IBD.[28] This finding was confirmed in this study. This may have been the result of the smaller amount of PSC patients without IBD included in this analysis, however, it could also suggests that the association between smoking and PSC is driven by a – specific – genotype and corresponding PSC-IBD phenotype. The underlying mechanisms through which smoking affects the pathogenesis of PSC or IBD are largely unknown. Tobacco contains more than 4000 chemicals, of which nicotine is studied the most.[29] Nicotine activates nicotinic acetylcholine receptors in the central nervous system, autonomic ganglia and neuromuscular junctions, but also on immune cells and in the mucosa and lymphoid tissue of small and large bowel.[29] van der Zanden et al.[30] showed that repeated exposure to nicotine or tobacco smoking elevates α7 nicotinic acetylcholine receptor expression on human monocytes which may lead to sensitization of the 'cholinergic anti-inflammatory pathway'. More recently, nicotine has been shown to elevate mRNA expression of the pro-fibrogenic factors collagen 1-α2 and TGF-β (transforming growth factor beta) and induces human hepatic stellate cell proliferation.[31] The effect of smoking on the progression of PSC has never been studied, yet in patients with primary biliary cirrhosis smoking has been associated with advanced histological disease at presentation.[32] These findings may suggest that smoking is associated with a lower risk of developing PSC, but once liver fibrosis is present smoking might accelerate disease progression.

Although several studies have shown a protective effect of appendectomy on developing UC, a meta-analysis combining 342 PSC patients from four studies showed no association with PSC.[21,23,24] This was in accordance with this study, where we observed more appendectomies in PSC-UC patients and HC than in UC controls. These findings match the previously described protective effect of appendectomy in UC. However, appendectomy does not seem to result in a lower risk of PSC, as there is no difference in frequency of appendectomy between the PSC-UC subgroup an HC. The higher frequency of appendectomy in the PSC-UC group compared to the UC group indicates that appendectomy may be a risk factor for PSC in UC patients. This observation could be explained by a distinct genotype and IBD phenotype of PSC-UC patients.[4] Possible immune-modulating effects of the appendix have been the focus of research for many years. A large study from Sweden performed between 1964 and 1993 including 212 963 patients showed that patients who underwent an appendectomy for an appendicitis or mesenteric lymphadenitis but not for non-specific abdominal pain before the age of 20 had a low risk of UC.[33] These findings suggest that the inflammation rather than the absence of the appendix has a protective effect. Although the human biological pathways involved in the protective effect of appendicitis or complete removal of the appendix are just starting to be unravelled, an appendicitis mouse model showed down-regulation of 14 genes including IBD-associated genes after appendectomy.[34]

Bergquist et al.[7] showed in another large Swedish study that first-degree family members of PSC patients have a 3.8-fold increased risk of PSC and a 3.3-fold increased risk of developing UC. A drawback of this study design is that relatives and friends accompanying PSC and IBD patients to the outpatient clinic were included as HC, and therefore we were unable to study the prevalence of PSC and IBD in families of healthy volunteers. In addition, this composition of the HC group could have introduced selection bias regarding the other environmental factors that were studied. On the other hand, using household members to some extent controls for unknown environmental factors, such as diet.

The balanced geographical distribution of PSC patients at time of diagnosis over cities and villages suggests that residential environment does not play a role in the aetiopathogenesis of PSC. In agreement with our findings, Ala et al.[35] found no association between standardized prevalence ratios of PSC patients listed for liver transplantation in New York City and proximity to super fund toxic waste sites.

This study has several limitations. First of all, cases and controls could unfortunately not be matched evenly for number of patients, age and gender. PSC patients were derived from 44 hospitals in the Netherlands, whereas control patients were derived from only four of these hospitals, explaining the relatively lower numbers. To rule out differences among subjects based on site of recruitment, a sensitivity analysis was performed including only cases and controls from the same site of recruitment. Observations remained consistent (Table S2 http://onlinelibrary.wiley.com/doi/10.1111/liv.12894/suppinfo). Because of the unbalanced gender and age distribution of PSC patients at diagnosis, one-on-one sex and age matching was not possible.[1] The differences in age and gender were accounted for in the multivariable regression analysis. Secondly, on average, three per cent of all IBD patients have concomitant PSC, 40% of these will develop their PSC after the diagnosis of IBD.[4] Therefore 1.2% of IBD controls may develop PSC in the future, which would disqualify them as proper controls. Lastly, several confounding factors may play a role when studying smoking behaviour. Suffering from a life threatening chronic illness such as PSC may influence tobacco use. Notably, controls were recruited from outpatient clinics, which could have influenced the number of smokers. However, 27% of the general population (>14 years of age) in the Netherlands was smoking in 2010, implying that the percentage of smokers at time of inclusion in the patient and control groups did not exceed the national average.[36] When interviewing patients about their past smoking behaviour, there is the risk of recall bias. By keeping questions on smoking behaviour simple, we attempted to minimalize this risk. The downside is that our questionnaire did not capture intermittent smokers or smokers who quit more than 20 years ago. In the general Dutch population, – studied since 1989 – the percentage of smokers is highest in the age group 25–44 years for males, and in the age group 16–24 years for females.[37] Thereafter, with each age increment, the percentage of smokers decreases.[37] Therefore, most PSC patients will have commenced smoking before the age of 44 (male) or 24 (female). As the median age at diagnosis of the PSC patients was 39 years (IQR 29–49) the vast majority of PSC patients will have started smoking before their PSC diagnosis.

We did not register the indication of appendectomy, and were not able to asses if inflammation rather than the absence of the appendix could be a risk factor for PSC, as was suggested for UC.[33] Given the high frequency of ileocecal resection in CD patients, appendectomies were not compared between the PSC-CD and CD subgroups. Only appendectomies performed before PSC diagnosis were included in the risk factor analysis, so IBD controls and HC's had more time at risk for undergoing appendectomy. This could have resulted in a somewhat higher number of appendectomies in the HC and IBD control groups compared with the PSC group, making our observation of relatively high appendectomy rate in the PSC groups even more notable.

In conclusion, we confirm in a large population-based cohort that smoking is associated with a lower risk of developing PSC, independent of the protective effect in UC. Appendectomy is not an independent risk factor for PSC.