Predicting Complicated Crohn's Disease and Surgery

Phenotypes, Genetics, Serology and Psychological Characteristics of a Population-Based Cohort

J. D. Ryan; M. S. Silverberg; W. Xu; L. A. Graff; L. E. Targownik; J. R. Walker; R. Carr; I. Clara; N. Miller; L. Rogala; C. N. Bernstein


Aliment Pharmacol Ther. 2013;38(3):274-283. 

In This Article


This study confirms that a significant relationship exists between stricturing and penetrating disease behaviour and the need for surgical intervention. At 5 years from inclusion in the Manitoba IBD Cohort Study and a median of 9 years from diagnosis, 65% of our subjects had complicated disease and 42% had undergone surgery. Disease location remained quite stable over this time span and disease behaviour progressed in a minority of subjects (Table 2). Subjects with stricturing or penetrating disease were on average six times more likely to undergo surgery, even after accounting for all other variables in our analysis.

The evolution of CD in our cohort is consistent with the majority of population-based studies to date. However, there is considerable methodological variation as well as differences in the incidence of surgery and disease complications reported in the literature.[1] At a median of 10 years from diagnosis, stricturing or penetrating complications occurred in 39% of subjects in Olmsted County,[43] 53% of a Norwegian inception cohort[44] and 56% of a New Zealand cohort.[45] In Olmsted County, 58% of subjects underwent surgery by a median of 13.2 years from diagnosis.[46] At 10 years from diagnosis, 38% of subjects in Norway[44] and 71% from Sweden[47] had undergone surgery. At 5 years from diagnosis, surgery had been performed in 39% of subjects from Wales[48] and 21% from Hungary.[49] Where reported, stricturing and penetrating disease behaviour has been a key factor associated with surgery.[43,44,48,49] Although some studies found ileal location to be associated with complicated disease or surgery,[43–45,49] others found an association with disease involving the colon.[48,49] Age at diagnosis was predictive of surgery in Norway[44] and Sweden,[47] although the associated ages were different in these two cohorts. Similar to our findings, perianal disease was associated with surgery in Sweden[47] and stricturing and penetrating complications in New Zealand.[45] In contrast to other population-based studies, we did not find any significant correlations between age at diagnosis and disease location with complicated CD or risk of surgery. However, as noted above, there are conflicting data as to which specific phenotypes are key predictors. Whether the absence of such correlations reflects our relatively small sample size or actual systematic differences between our population and others remains unclear. Nevertheless, CD characterised by stricturing and/or penetrating behaviour has consistently been associated with higher rates of surgery across populations and is likely the most important, independent phenotypic risk factor. We excluded complicated disease from our multivariate analysis of predictors of surgery as it was an independent outcome as well.

Previously established markers for CD analysed in our study cohort included positive ASCA IgA and IgG, mutations in NOD2, ATG16L1, IL23R, family history of IBD, Jewish ethnicity and tobacco use.[41,50,51] Both the number and magnitude of antibodies present in the sera of CD subjects have been shown to correlate with disease location, behaviour and the need for surgery independent of the NOD2 genotype.[52,53] However, some studies have suggested that CD subjects with NOD2 variants have a hypersensitive adaptive immunological response leading to increased antibody formation.[42,54] Therefore, it remains unclear whether the presence of specific antibodies or susceptibility genes are indeed independent risk factors for more complicated disease. The presence of ASCA IgG in our study was on average associated with a 3-fold increase in the risk of complicated CD and a 2-fold risk of surgery even after inclusion in the multivariate model. ASCA IgG was also associated with increased risk of disease confined to the terminal ileum. Another important finding of our study was that, in general, serological antibody testing for ASCA remained stable over the 5-year interval suggesting that serological status may be determined at any time with similar prognostic implications. Few population-based studies have so far confirmed a significant association between ASCA and complicated CD and our results represent an important contribution to this literature.[55]

No other antibodies proved to be significant predictors of complicated CD or surgery in our cohort. To date, the remainder of the antibodies tested in our study has been associated with complicated CD or surgery in at least one study, but results are conflicting.[56] Other studies have used quartile sum scores to demonstrate higher amalgamated antibody response in subjects with more complicated CD and higher rates of surgery.[52,53,57–59] Whether quartile sum scores provide additional prognostic information beyond basic ASCA serology remains unclear and requires further population-based testing. Our data suggest that knowledge of ASCA IgG status alone was sufficient in our population to predict worse disease outcomes.

At least 27 variants of NOD2 have been identified, but three mutations, R702W, G908R and 1007 fs, appear to account for the majority of variation seen in the CD population.[60] A recent analysis of 71 CD loci estimated that when combined they accounted for less than 25% of predicted heritability.[61] Hence, with our relatively small sample size, we focused on the more commonly associated SNPs. Other than the three NOD2 SNPs, we analysed for SNPs in IL23R and ATG16L1, which may be correlated with disease severity.[10–17] The only SNP to emerge as a suggestive predictor of complicated CD or surgery in our study was NOD2 rs2066847; however, this association did not remain significant after inclusion in the multivariate model. Elsewhere, NOD2 variants have been associated with complicated disease and surgery.[10,62] In a recent meta-analysis of the association between NOD2 mutations and CD outcomes, there was significant heterogeneity across all studies and the majority were referral centre-based.[10] Therefore, there remains a need for larger population-based studies to confirm the association between NOD2 variants and more complicated CD and surgery. We performed an additional analysis to look for correlations between SNPs and serological markers and found a trend towards an association between NOD2 rs2066847 and ASCA IgG (P = 0.1) and IgA (P = 0.06). Hence, NOD2 and ASCA may not be independently associated with disease activity and further studies are required to determine the degree of correlation between genetic and serological markers.

Similar to the results of a population-based study from New Zealand,[14] we failed to demonstrate a significant relationship between SNPs in ATG16L1 and IL23R and any outcomes of interest. This may reflect further differences between subjects with CD presenting to tertiary referral centres and those in the general population. Alternatively, the influence of ATG16L1 and IL23R on phenotype may be modest, requiring much larger sample sizes to reach statistical significance. Our study suggests that identifying common NOD2 variants alone may be helpful in further delineating those subjects at risk of more complicated CD and surgery, but that this information may be of little additional benefit in the presence of other phenotypic or serological risk factors. We also found no association between sex, smoking status or family history and complicated CD or risk of surgery.

Regarding smoking, we were unable to demonstrate a significant risk of complicated CD or surgery. Instead, we report a weak association between smoking and inflammatory disease behaviour without strictures or fistulae. We also found that smokers were less likely to experience perianal disease behaviour. Most evidence for smoking as a risk factor comes from referral centre studies.[63,64]Whether our results are spurious or underscore, an actual difference between subjects attending referral centres and those in general population will require more population-based studies to elucidate.

Finally, we undertook what we believe to be the first study of psychological variables in association with CD phenotype, genetic and serological markers. Psychological variables were included as they have been found by our group and others to be associated with disease activity. We have reported on perceived stress and distress in CD and their association with active disease.[32,33] We have also reported that these subjects are more likely to have depression diagnosed years prior to their IBD diagnosis.[34] Furthermore, we have reported that a high level of perceived stress was the only statistically significant predictor of a symptomatic flare of CD activity in a subsequent 3-month period.[65] So, although psychological variables may influence future IBD diagnosis and disease activity, it is unknown if they influence CD phenotype or outcome. Although psychological characteristics did not appear to be significant predictors of CD outcomes when included in the multivariate analysis, we did find some associations with potentially important clinical implications that warrant further exploration. Our results suggest that many of our subjects with CD have experienced adversity in childhood. There were trends of associations between surgery and higher childhood adversity scores, and in particular with childhood physical abuse as one form of adversity. Undoubtedly, adverse childhood experiences occurred long before disease was presented in our cohort who were for the most part diagnosed after the age of 17. This finding may underscore the need for further exploration as to how early childhood stressors may set the stage for later intestinal inflammation. Taken together, these findings suggest that clinicians should anticipate that those subjects undergoing surgery for CD may be particularly vulnerable to increased stress during the course of their illness. Therefore, we feel that it is important to consider the potential interaction between psychological characteristics and disease outcomes when considering the long-term prognosis of CD in the general population.

The strength of our study lies in the fact that it is population-based and hence less likely to suffer referral centre bias. A referral centre bias may lead to overestimates of the effects of risk factors by focusing primarily subjects with more aggressive phenotypes. Conversely, this bias may also underestimate the magnitude of risk factors by failing to include sufficient numbers of subjects with less aggressive disease as comparators. Our study design also attempted to control for lag time bias using a comprehensive update of our subjects' phenotype and disease course over a 5-year interval. By reporting phenotype at 3 years from diagnosis and again at 5 years from inclusion, we allowed time for adequate investigation to be undertaken and for subjects to manifest their CD phenotype. We are reporting associations with the important outcomes of complicated disease or surgery at a median of 9.3 years of disease duration. However, a weakness of our study is that we did not have specific time points at which participants had evaluations of their phenotype or surgery status. Hence, a survival analysis showing the evolution of phenotype, for instance, over time could not be undertaken. The most important weakness of our study is the relatively small sample size, especially for genetic analysis. For this reason, we limited our analysis on six SNPs in three major CD susceptibility genes, namely NOD2, IL23R and ATG16L1, which were previously established correlates of CD in our cohort.[41] Nevertheless, our study was underpowered to detect all but the most strongly correlated factors.

In summary, from a clinical perspective, the usefulness of any disease marker or classification scheme will be a measure of its ability to predict the natural history of disease and to direct therapy accordingly. On the whole, the results of our study reinforce several important findings with regard to the prognosis of CD. In our subjects, who were followed up for every 6 months over a 5-year period, we report the following: (i) more than half of subjects will experience a significant complication from their disease within the first 9 years from diagnosis; (ii) disease behaviour is the single greatest predictor of surgical risk, especially early surgery; (iii) ASCA IgG is an independent predictor of complicated CD and surgical risk; and (iv) while there are some associations with NOD2 risk alleles, they may not be independent of risks associated with ASCA, which currently is an easier test to access. For clinicians, our results should emphasise the importance of classifying CD phenotype early in the course of disease when most disease morbidity is likely to be encountered and the potential to identify those subjects at risk of surgical intervention is greatest. Our study also provides a rationale for undertaking serological testing for ASCA IgG as a means of not only discriminating CD from ulcerative colitis but also improving prognostication. This may be of particular importance to clinicians who encounter CD earlier in its natural history and who are increasingly asked to balance the benefits of early, aggressive intervention against the long-term risks and costs associated with this strategy. Our results suggest that ASCA IgG is the most important marker in serological panels for assessing prognosis in CD. In addition, the CD-associated risk variants assessed here do not seem to add value as predictors of CD prognosis. Ultimately, it is imperative that any classification scheme and any proposed new marker of CD be tested in population-based cohorts where clinical utility will determine the relevance to everyday practice.