Estimating the Risk of Ulcerative Colitis-Associated Pyoderma Gangrenosum: A Population-Based Case- Control Study


 BackgroundUlcerative colitis (UC) is a well-known underlying comorbidity in pyoderma gangrenosum (PG). However, the relative risk conferred by UC for the subsequent development of PG is yet to be explicated.ObjectiveTo estimate the magnitude of the association between UC and the subsequent occurrence of PG, thus enabling to assess the PG risk with UC.MethodsA population-based case-control study was conducted comparing PG patients (n = 302) with age-, sex- and ethnicity-matched control subjects (n = 1,497) regarding the presence of UC. Logistic regression models were utilized for univariate and multivariate analyses.ResultsThe prevalence of UC was greater in patients with PG than in control subjects (7.3% vs. 0.5%, respectively; P < 0.001). A 15-fold increase in the odds of PG with UC (OR, 14.62 95% CI, 6.45–33.18) was observed. This association retained its statistical significance following a sensitivity analysis excluding UC cases diagnosed up to 2 years prior to PG (OR, 9.88; 95% CI, 3.91–24.97), and after adjusting for confounding factors (adjusted OR, 10.78; 95% CI, 4.55–25.52). The median latency between the diagnosis of UC and the development of PG was 4.59 years. Patients with both PG and UC were younger and had a lower prevalence of smoking when compared to the remaining patients with PG.ConclusionsUC increases the odds of developing PG by 15-folds. Physicians managing patients with UC should be aware of this increased burden. Patients with UC may be advised to avoid additional precipitating factors of PG.


Abstract Background
Ulcerative colitis (UC) is a well-known underlying comorbidity in pyoderma gangrenosum (PG). However, the relative risk conferred by UC for the subsequent development of PG is yet to be explicated.

Objective
To estimate the magnitude of the association between UC and the subsequent occurrence of PG, thus enabling to assess the PG risk with UC.

Methods
A population-based case-control study was conducted comparing PG patients (n = 302) with age-, sex-and ethnicity-matched control subjects (n = 1,497) regarding the presence of UC. Logistic regression models were utilized for univariate and multivariate analyses.

Results
The prevalence of UC was greater in patients with PG than in control subjects (7.3% vs. 0.5%, respectively; P < 0.001). A 15-fold increase in the odds of PG with UC (OR, 14.62 95% CI, 6.45-33. 18) was observed. This association retained its statistical signi cance following a sensitivity analysis excluding UC cases diagnosed up to 2 years prior to PG (OR, 9.88; 95% CI, 3.91-24.97), and after adjusting for confounding factors (adjusted OR, 10.78; 95% CI, 4.55-25.52). The median latency between the diagnosis of UC and the development of PG was 4.59 years. Patients with both PG and UC were younger and had a lower prevalence of smoking when compared to the remaining patients with PG.

Conclusions
UC increases the odds of developing PG by 15-folds. Physicians managing patients with UC should be aware of this increased burden. Patients with UC may be advised to avoid additional precipitating factors of PG.

Background
Pyoderma gangrenosum (PG) is a rare neutrophilic in ammatory dermatosis that typically presents with one or more painful ulcers with violaceous, undermined borders on the lower extremities. Less common morphological phenotypes include pustules or tender nodules on other anatomical locations [1]. Histologically, it is typi ed by a neutrophil-rich in ltrate. PG may be sporadic or present in syndromic forms, where it is regarded as a polygenic autoin ammatory disease [2,3]. More than half of patients with sporadic PG have an underlying systemic disease, mainly in ammatory bowel disease (IBD), rheumatoid arthritis, and hematological malignancies [4].
Ulcerative colitis (UC) is a chronic idiopathic in ammatory bowel disorder manifesting with continuous mucosal in ammation extending from the rectum to the more proximal portions of the colon. The disease is characterized by a relapsing and remitting course imposing a high burden of morbidly and substantial disruption in quality of life [5]. While the incidence and prevalence of UC demonstrate a decreasing or stabilizing trend in Europe and North America during the past three decades, its incidence has been increasing in newly industrialized countries in Asia, Africa, and South America [6]. UC may be complicated with extraintestinal manifestation involving musculoskeletal, cutaneous, ocular, and hepatobiliary conditions [7].
UC was long considered as one of the leading underlying comorbidity among patients with PG. The pooled prevalence of UC among patients with PG had been estimated at 11.5% (95% CI, 7.2-16.6; I2 = 83.7%) in a recent meta-analysis synthesizing data across 17 cohorts of patients with PG [4]. Correspondingly, the incidence of PG among patients with IBD was estimated at 0.6% in a meta-analysis summarizing cohorts of IBD patients [8]. Nonetheless, patients with PG were not compared with matched controls with regard to the presence of UC. Thus, the relative risk of developing PG after the occurrence of UC is yet to be outlined.
The aim of the current study is to quantify the association between previous UC and the later occurrence of PG in a population-based case-control study. Additionally, we aimed to evaluate whether patients with comorbid PG and UC have distinct epidemiological features distinguishing them from other patients with PG.

Results
The study population comprised of 302 patients with PG and 1497 matched control individuals. The mean (SD) age at the diagnosis of patients and enrollment of control individuals was 54.0 (20.8) years. Onehundred-twenty-seven (42.1%) patients with PG were males, and 255 (84.4%) were of Jewish ancestry. The sex distribution and ethnic background were comparable between the two groups (Table 1). Moreover, the average BMI and the prevalence of smoking were not signi cantly different in the two groups ( Table 1). The mean (SD) Charlson comorbidity score was higher in case than in control individuals ( Table 2).   (Table 4).
Patients with both conditions had a lower prevalence of smoking (13.6% vs. 40.0%, respectively; P = 0.001), while the mean BMI was comparable between the two subgroups (Table 4)

Interpretations of the current study`s main observations
Despite the aforementioned data, the magnitude of the association between UC and the subsequent occurrence of PG is yet to be delineated. This gap in our knowledge rises from the lack of controlled studies comparing patients with PG and controls regarding the prevalence of preceding UC. Therefore, the current case-control study provides a novel epidemiological feature and conveys the important message that a foregoing diagnosis of UC is implicated with an 11-and 15-folds increased multivariate and crude odds of having PG, respectively. In other words, individuals with UC have 11-to 15-fold greater odds to present with PG as compared to individuals without a diagnosis of UC. This conclusion is grounded upon the 'rare disease assumption' in case-control studies, theorizing that the OR approximates the RR in rare diseases [11]. Patients with UC should be aware of this substantial risk in order to minimize other risk factors of PG, including pathergy phenomenon derived from piercing, trauma, or unnecessary surgical procedures. Awareness of this risk may be of great bene t for physicians encountering rapidly evolving ulcers in patients with UC.
We found an explicit predilection for PG among female patients. Previous studies demonstrated that females were more predisposed to experience extraintestinal manifestations of IBD. In the meta-analysis of States et al. [8], female sex imposed a 32% increased risk of developing PG among patients with IBD (RR, 1.33; 95% CI 1. 16-1.52). In addition, Weizman et al. [12] noted a female preponderance (58%) among their 80 patients with IBD-associated PG. In a Swedish population-based study including 1,274 patients with UC, Monsen et al. [13] observed a threefold increased odds of erythema nodosum among females. In their nationwide study of 950 Swiss patients with IBD, Vavricka et al. [14] found that extraintestinal manifestations were more prevalent among female patients with IBD (50% vs. 34% in males). The aforementioned observations may imply for a putative role of estrogen, gene-hormone and/or gene-gene interactions in the induction pathogenesis of PG and other extraintestinal UC manifestations of UC. Further research is required to spell out sex differences in PG.
Our study also denotes that patients with coexistent PG and UC were signi cantly younger at the onset of PG relative to their remaining PG counterparts. The main interpretation of this nding stems from the fact that other patients with PG assumingly include a sizable proportion of patients with underlying hematological malignancies and rheumatoid arthritis who tend present at older ages, thus delaying the development of secondary PG. This nding aligns with previous studies tracking patients with IBDassociated PG, which reported younger ages of presentation [12,15].
We revealed signi cantly greater odds of PG among patients of Arab ethnicity. The existence of ethnic predisposition in PG is still a debatable and inconclusive issue. While Nguyen et al. [16] did not nd any ethnic predisposition for cutaneous manifestation in IBD, Farhi et al. [17] reported a higher risk of PG among IBD patients of an African race. Of note, both the incidence and prevalence rates of IBD were found lower among the Arab population residing in Israel relative to their Ashkenazi Jewish counterparts [18]. However, the Arab minority in Israel is typi ed by a higher burden of IBD when compared to Arab populations originating from other Middle-Eastern countries [19,20]. In a study aiming to address morphological differences between IBD patients from different ethnic groups in Israel, patients of Arab ancestry were found to present more frequently with extraintestinal manifestations [18]. The latter nding lends credibility for the results of the current study.
The association between UC and PG was more protruding among non-smokers. This nding accords with the conclusions drawn from several observational studies suggesting that cigarette smoking does not confer risk for cutaneous manifestations in IBD (including both PG and erythema nodosum) [8,21,22]. In contrast, cigarette smoking was a signi cant predictor of PG in a cohort of patients with UC [23]. It is noteworthy that cigarette smoking emerged as a protective factor against the development of UC [24]. A meta-analysis summarizing 13 epidemiological studies has con rmed that current smoking imposes a protective effect in UC, with smokers having an OR of 0.58 (95% CI, 0.45-0.75) to develop UC as compared to non-smokers [25]. Further research is required to better explicate the precise role of smoking in the development of cutaneous manifestation in UC.

Strengths and limitations
The case-control design enables the identi cation of the temporal relationship between diagnoses and allows the estimation of a causal relationship between the two conditions of interest [26]. The populationbased design and the recruitment of cases managed both in inpatient and outpatient settings diminish the likelihood of selection bias that may arise in other observational studies.
One of the main drawbacks interfering with the current study is the lack of data concerning the clinical characteristics and severity indices of the two diseases. Although the formal disease criteria were not directly required to recruit patients, the inclusion criteria were strict and valid; based on documentation by a certi ed dermatologist or dermatological wards for PG and on the chronic disease registry of CHS for UC. The latter was proven highly reliable [9].

Conclusions
In conclusion, the current population-based case-control study depicts that the presence of UC is associated with 15-fold increased odds of having subsequent PG. The association was more protuberant among female, younger, and Arab patients, as well as among non-smokers. Estimates were not signi cantly altered in a sensitivity analysis excluding diagnoses up to 2 years before PG and in multivariate analysis adjusting for confounding factors.

Study design and database
We performed a retrospective population-based case-control study aiming to identify the risk of having subsequent PG following the diagnosis of UC.
The computerized dataset of Clalit Health Services (CHS) was the origin of the current study. CHS is the largest health care organization in Israel, providing healthcare services for approximately 4,400,000 enrollees. It provides a wide array of both public and private healthcare services. The computerized database of CHS is characterized by incessant real-time input from medical, administrative, and pharmaceutical operating systems. It was proven highly competent to enable the performance of epidemiological studies.
The chronic disease registry of CHS retrieves data from various sources, including from hospital and primary care reports, drug claims. These accumulated data is then manually cross-checked and validated by the primary care practitioner of each patient. This registry was considered as of vast reliability on the grounds of previous epidemiological studies [9].
The ethical committee of Ben-Gurion University approved the current study which was performed in accordance with declaration of Helsinki.

Study population and covariates
Between the years 2000 and 2018, we identi ed all individuals with diagnostic codes consistent with the diagnosis of PG. The inclusion of PG cases in this study was based on the following criteria: (i) a documented diagnosis of PG at least twice in the medical le as registered by a community board-certi ed dermatologist, or (ii) documentation of a diagnosis of PG in hospital discharge letters from dermatological wards.
The diagnosis of UC was based on its registration in the chronic registry of CHS. To elaborate, it is based on the documentation of board-certi ed gastroenterologist, on the claims of UC-related drugs, and is eventually validated by the managing general practitioner.
The control group included up to 5 controls per patient, matched randomly by age, sex, and ethnicity. Age matching was based on the exact year of birth (1-year strata). The diagnosis date was used as an index date for the cases, and each matched control patient. Controls were ascertained to be alive and to contribute longitudinal data to the CHS dataset at their recruitment date.
Outcome measures were adjusted for comorbid underlying diseases utilizing the Charlson comorbidity index, a validated epidemiological method of measuring and categorizing comorbidities. This index has been demonstrated to be reliable in forecasting mortality [10]. Smoking status was classi ed as a current smoker or never/past smoker.

Statistical analysis
Baseline characteristics are described with means and standard deviations (SD)s for continuous variables, whereas percentages are used to represent categorical values. A comparison of the distribution of sociodemographic and clinical factors between patients with PG and controls was conducted using the Chi-square test and t-test, as indicated.
Logistic regression was used to calculate odds ratios (ORs) and 95% CIs to compare cases and control with respect to the presence of UC. The association was calculated based on individuals who developed PG following the diagnosis of UC in accordance with the presence of a temporal relationship between exposure and outcome in case-control studies. Two-tailed P-values less than 0.05 were considered as statistically signi cant. All statistical analyses were performed using SPSS software, version 25 (SPSS, Armonk, NY: IBM Corp).
A sensitivity analysis was performed to ensure that the observed association was not overestimated due to ascertainment bias. This analysis was based on repeating all calculations with an exclusion period of two years prior to the diagnosis of PG. Thus, all individuals given a diagnosis of UC in this period were omitted from all calculations. Ethics approval and consent to participate

List Of Abbreviations
The current study was approved by the institutional ethical committee of Ben-Gurion University (0212-17-COM). Informed consent was deemed unnecessary according to local regulations since this is a retrospective non-interventional study.

Consent for publication
Not applicable