Perianal Crohn’s Disease: Predictive Factors and Genotype-Phenotype Correlations

Identification of Crohn’s disease (CD) patients who would or would not develop perianal CD (PCD) could broaden surgical options in these patients. Knowing that a patient would not develop PCD could for example perhaps allow them to undergo sphincter-sparing surgery such as ileal pouch anal anastomosis. While there are many reports on the genetics and clinical progression of CD, few offer insight on predisposing factors to PCD, a phenotype of the disease that is a source of special morbidity in these patients. Perianal disease has been reported in 13–43% of patients with CD [1,2] and is one of the major factors preventing sphincter-sparing surgery in these patients. Perianal manifestations may include skin tags that pose hygiene problems, anal fissures and ulcers, anal canal strictures, perianal fistulae, and abscesses [3]. Treatment of PCD is complicated both by the manifestations themselves and their location; frequent bowel movements prevent proper healing and increase chances of local sepsis. Uncontrollable fecal or purulent drainage can lead to skin maceration and urinary tract infections, especially in women with rectovaginal fistulae. Patients with PCD endure not only the debilitating complications from their disease, which often fail medical therapy and require permanent fecal diversion, but also the strain it places on their social and sexual lives [4]. As iterated above, identification of CD patients likely to develop PCD would have several clinical advantages. Patients thought not to be at risk for developing PCD could be selected for sphincter-sparing surgery, while patients at higher risk for developing PCD could perhaps be placed on some type of chemoprophylaxis to prevent PCD and its associated tissue and sphincter destruction.

Single-nucleotide polymorphisms (SNPs) in the interleukin-23 receptor (IL-23r) gene and the caspase recruitment domain 15 (CARD15/NOD2) gene correlate with increased susceptibility of acquiring CD [5,6]. Additional mutations within the organic cation/carnitine transporter (OCTN) gene and intergenic region (IGR) of the IBD5 gene cluster have been found to predict CD [7,8]. In addition, several reports have identified clinical factors that may predict a more aggressive course of CD including a family history of inflammatory bowel disease (IBD) and active cigarette smoking, a young age at onset, and a diagnosis of CD in the ileum [9,10,11].

Few reports have studied genotype-phenotype correlations that may specifically predispose CD patients to PCD. An association was previously identified between a genetic variation in the IBD5 haplotype and CD; furthermore, IBD5 variants were most strongly associated with perianal complications of CD [12,13]. Two years later, Vermiere et al. [14]proposed that variants in the OCTN1 gene, which is located within the IBD5 gene cluster, are involved in the phenotypic expression of PCD, but not in CD susceptibility itself. Freire et al. [15] noted that CARD15/NOD2 mutations did not predispose to perianal fistulization. Similarly, Karban et al. [16]did not find CARD15/NOD2 or OCTN to be a risk factor for PCD in a cohort of non-Ashkenazi Jews with CD.

Based on the above, we examined genotype-phenotype correlations as associated with the occurrence of PCD with the hope of determining predictive criteria that could be used prospectively in clinical patient care.

This study was approved by the University of Louisville Institutional Review Board, and written informed consent was obtained from all subjects. Clinical and genetic data were prospectively entered into a password-protected Microsoft Access relational database containing data on over 3,000 medically and/or surgically treated digestive disease patients treated between 1997 and 2008 [5].

The diagnosis of CD was made via conventional endoscopic, radiologic, and histological findings. In cases of colonic CD, the diagnosis was verified by a single gastrointestinal pathologist with a special interest in IBD. For purposes of our study, PCD was defined as presence of at least one of the following: perianal fistulae and/or abscesses, anal canal fissures or ulcers, and anal canal stenosis. Controls were comparison patients seen in the same practice who presented for evaluation of noninflammatory, nonmalignant conditions (internal hemorrhoids, constipation, etc.).

Pertinent clinical information for each patient was carefully reviewed, including cigarette smoking status using the Behavioral Risk Factor Surveillance Survey [17]; family history of IBD; family history of colorectal cancer; number of operations for CD; operations for PCD, including the need for permanent fecal diversion; and last date of follow-up. Patients for whom long-term follow-up was not achieved (defined as not having been seen for >1 year; n = 158) were excluded from this study. Each patient was classified according to the Vienna classification with respect to age at onset of disease (≤40 years, >40 years), location of disease [terminal ileum (L1), colon (L2), ileocolonic (L3), upper GI (L4)], and disease behavior [inflammatory (B1), stricturing (B2), penetrating (B3)] [18].

Genomic DNA was extracted from peripheral whole blood with a Puregene® DNA extraction kit (Gentra Systems, Inc., Minneapolis, Minn., USA). Polymerase chain reaction (PCR) amplification and genotyping were performed using an ABI prism 7300 sequence detection system (Applied Biosystems®, Foster City, Calif., USA). SNP-specific TaqMan® PCR primers and fluorogenic probes were obtained from Applied Biosystems. The fluorogenic minor groove binder TaqMan probes were labeled with a reporter dye specific for the wild-type and variant alleles of eleven IBD-associated SNPs.

Eleven SNPs (table 1) chosen for their reported association with CD and/or PCD were genotyped for all CD and PCD patients as well as controls.

Descriptive statistics related to genetic and clinical characteristics were produced for all patients. Categorical variables were compared using the Pearson χ² test (or Fisher’s exact test) for contingency tables [19]. The t test or Wilcoxon rank sum test was used to examine the cohort for continuous variables. We also fit the univariable and multivariable logistic regression models for the probabilities of patients in PCD or CD groups regarding their possible genetic and clinical predictors [20].

Hardy-Weinberg equilibrium and genotype and allelic association for 11 SNPs were assessed using the Pearson χ² test. For allelic association analysis, we also used logistic regression to estimate odds ratios and 95% confidence intervals for patients with PCD or non-PCD as compared to the control group, or for patients with PCD as compared to patients with non-PCD. Logistic regression was also used to determine the interaction effects with inclusion of the main effects of each gene in the model. Finally, haplotype analysis was conducted and NOD-2 (3 SNPs), IL-23r (5 SNPs) and IGR(2 SNPs) were compared between PCD and non-PCD patients. All calculations were performed with SAS statistical software [21].

The study cohort consisted of a well-defined patient population originating from a small, largely rural geographic area in the Midwest United States (Kentucky and southern Indiana). For purposes of genetic homogeneity, only Caucasian patients were included in this analysis. The Caucasian cohort contains 795 individuals: 317 patients with CD (170 of whom have CD without perianal disease [non-PCD] 39% male) and 147 with PCD (36% male), and 478 controls (32% male; table 2). The mean patient age was 51 years for non-PCD, 48 years for PCD, and 56 years for controls (p > 0.05). Forty-four percent (n = 141) of all CD patients in our cohort, both PCD and non-PCD, had a positive family history of IBD, defined as a first- or second-degree relative with IBD, as compared to only 9% of controls (n = 42).

Of the non-PCD patients, 30% were current smokers, 31% are former smokers, and 39% have never smoked (n = 50, 53, and 67, respectively). Of the PCD patients, 23% were current smokers, 34% former smokers, and 43% have never smoked (n = 34, 50, and 63, respectively). In our control group, 22% were current smokers, 34% formerly smoked, and the remaining 44% have never smoked (n = 92, 138, and 182, respectively; p > 0.05).

There was no significant difference in the age at diagnosis between the PCD and the non-PCD group, 29 versus 33 years, respectively (p > 0.05). Only 25% of non-PCD patients underwent more than three operations for their disease at the time of last follow-up (n = 43), while 52% of the PCD group had had more than three operations for their disease at the time of last follow-up (n = 76; p < 0.05).

Patients with PCD were more likely to have disease in the colon and ileocolic regions (Vienna L2, L3; 116 PCD vs. 96 non-PCD; 79 vs. 57%; p < 0.001; tables 3, 4), whereas patients with non-PCD were more likely to have inflammation in the terminal ileum and upper GI tract (Vienna L1, L4; 73 non-PCD vs. 31 PCD; 43 vs. 21%). Fifty patients with PCD required a permanent ileostomy (34%) as compared to only 6 non-PCD patients (4%).

On multivariate regression analysis, the number of operations and obviously the need for an ileostomy were both predictive of PCD. In contrast, CD located in the ileum/upper gastrointestinal tract (L1 + L4) or presence of inflammatory or stricturing disease behavior (B1 + B2) was more predictive of non-PCD (table 4).

For the genetic analyses, we considered only data from patients who had successful genotyping of at least 7 of the 11 SNPs. Overall, we analyzed data for 637 patients (170 CD, 147 PCD, and 320 controls). The control group was in Hardy-Weinberg equilibrium for all 11 SNPs. Univariate analysis of SNPs revealed the minor allele of each NOD2 SNP (1007insC, R702W, G908R) and two IL-23r SNPs (rs1004819, rs11465804) to be significantly associated with CD (p = 0.001). Additionally, two IL-23R SNPs (rs10489629, rs2201841) were found to be associated with PCD patients, but not the CD-only patients (table 5).

Minor allele expression in all SNPs was not significantly different between CD and PCD patients. Moreover, haplotype analysis of NOD2(3 SNPs), IL-23r(5 SNPs) and IGR (2 SNPs) genes revealed no significant difference in haplotypes between CD and PCD patients beyond two rare haplotypes of all five IL-23r loci.

A disease of unknown etiology, Crohn’s disease most likely describes several discrete diseases of the GI tract where inflammation arises due to currently unknown environmental antigens in the setting of a genetically susceptible host [22]. Tobacco use is one confirmed environmental factor that increases the severity of CD [22,23]. The risk for CD is 13-fold higher in first-degree relatives of individuals affected with CD [24], and CD concordance rates are 50–60% in monozygotic twins [25,26]. Taken together, this suggests that CD progression is most likely coordinated by a combination of genetic and environmental factors.

Since CD is a complex disorder with several distinct phenotypes, it is thought that these different phenotypes might have unique risk factors and genetic identities. We wished to carefully survey clinical factors and genotype-phenotype correlations that might predispose a patient to PCD. Such an association could have important implications for patient treatment in that identification of patients not destined to develop PCD could permit patients to be considered for sphincter-sparing surgery. Identification of patients with Crohn’s colitis not likely to develop PCD could permit some of these patients to undergo ileal pouch anal anastomosis. Many patients would choose this option over ileostomy even at the risk of subsequent inflammatory disease. There have been several reports of satisfactory control of inflammatory pouch-related symptomatology with agents such as infliximab [27,28]. PCD complications such as fistulae associated with stenosis are much more difficult to treat in this manner.

Studies contrast regarding gender as a risk factor for PCD. In one report of 221 CD patients with anal fistulae, there were 125 men (33%) and 96 women (22%), suggesting that men were more prone to PCD [29]. Conversely, our study and that of Schwartz et al. [30] found no association between gender and PCD. Differences in the overall gender distribution of the population samples studied and varying definitions for PCD may contribute to these conflicting reports on gender as a risk factor for PCD.

Family history of IBD is a strong predictor of CD behavior [9,31], but Dorn et al. [32] and Cosnes et al. [23] found no evidence that family history of IBD can be used to predict PCD. In our cohort, 77 non-PCD patients (45%) had a family history of IBD (defined as having a first- or second-degree relative with IBD) as compared to 42 controls (9%). This did not differ between non-PCD and PCD patients [77 patients (45%) and 64 patients (44%), respectively]. We do not have data regarding the incidence of PCD in family members with IBD. It is possible that family history of PCD could be a better predictor of PCD than family history of IBD alone.

Several reports indicate that patients who are diagnosed with CD at an early age are more predisposed to aggressive disease, including PCD [4,33]. Schwartz et al. [30] reported that 50% of CD patients who had CD for over 20 years developed fistulae, 26% of which were perianal, emphasizing the progressive nature of CD. Our data showed no difference in age at diagnosis between PCD and non-PCD patients (29 vs. 33 years, respectively, p > 0.05).

Multiple reports indicate that CD patients with colitis, ileocolitis, or proctitis are at a higher risk of developing perianal complications [4,16]. Not surprisingly, our data support this finding; according to the Vienna classification system [18], CD was located in the colon and ileocolon in 79% of PCD patients (n = 116), as compared to 57% of patients without PCD (n = 96). Colon/ileocolic CD location is a predictor of PCD (p < 0.001). Patients with colonic CD seem to be at higher risk of developing PCD.

Information regarding use of biologics was available in 44 CD patients (14%). Infliximab received approval for CD in the United States in 1998, adalimumab in 2007, and certolizumab in 2008. Infliximab was used in approximately 13.5% of CD cases where information regarding treatment was available (n = 23) and 14.3% of PCD cases (n = 21). Anti-TNF therapy, a chemoprophylaxis agent, is often used for long-term symptom control in PCD patients unresponsive to antibiotics and antimetabolites [34,35]. Prospective randomized controlled trials have shown that infliximab administered as a triple-dose induction regimen at 0, 2, and 6 weeks is associated with at least a 50% decrease in fistula drainage [3,36]. A previous report from our unit has, however, highlighted the failure even of biologic therapy in the presence of anal canal stenosis and extensive colonic disease [2]. Based on our findings, clinical predictors could be used to identify patients at higher risk for developing PCD. For example, patients with ileocolic or colonic disease would routinely be placed on 5-ASA or other medications as chemoprophylaxis to hopefully deter the development of PCD.

Our study confirms previous reports implicating CARD15/NOD2 and IL-23r SNPs as risk factors for CD [5,6]; however, we found no relationship between minor alleles in the CARD15/NOD2 and IL-23r genes and the incidence of PCD, coinciding with previous work by Freire et al. [15] and Karban et al. [16].

Also, in agreement with others [37], we did not identify OCTN1 and IGR variation within the IBD5 haplotype to be predictors of PCD. This is in contrast to the group of Armuzzi and Vermeire et al. [14] who did find OCTN1 and IGR to be predictors of PCD.

In viewing our data, the nature of our patient population should be considered. Unlike many other cohorts, this is largely a group of patients that has failed medical therapy and is referred for surgical management, thus selection bias may play a role. All PCD patients had failed medical therapy and slightly over half had had ≥3 operations, while 34% required a permanent ileostomy (n = 50). In addition, there is a profound environmental effect, as the population is derived from one of the areas in the United States with the highest rates of adult cigarette smoking [38]. Fifty-seven percent of PCD patients (n = 84) and 61% of non-PCD patients (n = 103) were former or current smokers. Our rural geographic area is one in which there is an unusually high rate of IBD. Our results may not be translatable to other populations of Crohn’s patients.

Colon/ileocolic CD location is a significant predictor of PCD. We found no genetic variations in the CARD15/NOD2, IL-23r, and OCTN1 genes or IGR that could help further identify patients at risk for PCD. More genotype-phenotype studies with large and well-characterized patient populations are needed to further explore clinical and genetic factors that predispose CD patients to PCD.

This publication was made possible by the John and Caroline Price Family Trust and a gracious gift from Ms. Sara Shallenberger Brown. Additionally, this work was supported by NIH/NIEHS grant 1P30ES014443-01A1 to The Center for Environmental Genomics and Integrative Biology. This work was also supported in part by the Wendell Cherry Endowed Fund, James Graham Brown Cancer Center, University of Louisville.

None of the authors has any financial or other conflicts of interest to disclose.