Clinical factors associated with severity in patients with inflammatory bowel disease in Brazil based on 2-year national registry data from GEDIIB

The Brazilian Organization for Crohn's Disease and Colitis (GEDIIB) established a national registry of inflammatory bowel disease (IBD). The aim of the study was to identify clinical factors associated with disease severity in IBD patients in Brazil. A population-based risk model aimed at stratifying the severity of IBD based on previous hospitalization, use of biologics, and need for surgery for ulcerative colitis (UC) and Crohn’s Disease (CD) and on previous complications for CD. A total of 1179 patients (34.4 ± 14.7y; females 59%) were included: 46.6% with UC, 44.2% with CD, and 0.9% with unclassified IBD (IBD-U). The time from the beginning of the symptoms to diagnosis was 3.85y. In CD, 41.2% of patients presented with ileocolic disease, 32% inflammatory behavior, and 15.5% perianal disease. In UC, 46.3% presented with extensive colitis. Regarding treatment, 68.1%, 67%, and 47.6% received biological therapy, salicylates and immunosuppressors, respectively. Severe disease was associated with the presence of extensive colitis, EIM, male, comorbidities, and familial history of colorectal cancer in patients with UC. The presence of Montreal B2 and B3 behaviors, colonic location, and EIM were associated with CD severity. In conclusion, disease severity was associated with younger age, greater disease extent, and the presence of rheumatic EIM.


Clinical and sociodemographic variables
The variables from the registration data included date of birth; current age; type of healthcare service utilized; ethnicity (as personally declared); sex; education; medical history of comorbidities; smoking status (never smoked, current smoker, or past smoker); city/state of outpatient care; type of IBD (CD, UC, or IBD-U); age at symptom onset, diagnosis, and registration; Montreal classification 14 ; disease severity; initial manifestations; extraintestinal manifestations (EIMs); presence of fistulas; comorbidities; and body mass index (BMI).Surgeries related to IBD, such as intestinal resections, placement of setons, colectomy due to intractable disease activity or neoplasia, and surgical complications (including complications related to short bowel syndrome or fecal incontinence) were also included.Information about previous hospitalizations (related to disease activity or IBD complications such as infectious diseases) and previous or current use of medications were recorded.In addition, data on family history of IBD or other immune-mediated diseases/cancers, including the degree of kinship, were also collected.

Outcome variables
Disease severity was characterized by previous hospitalization, use of biologics, and need for surgery in patients with CD and UC.Complications such as disease activity, pancreatitis, or infectious diseases (herpes simplex or zoster, tuberculosis, upper airway infection, fungal infection, pneumonia, and urinary infection) were also considered for patients with CD.For CD, the independent variables for the final model were the Montreal Classification (L-disease location, B-disease behavior), age at diagnosis (1-20 y, 21-40 y, 41-60 y, and 61-88 y), sex (male vs. female), smoking status (current vs. past), comorbidities (yes vs. no), and rheumatic EIMs (yes vs. no).For UC, the independent variables were the extent of the disease, age at diagnosis (1-20 y, 21-40 y, 41-60 y, and 61-88 y), sex (male vs. female), smoking status (current vs. past), comorbidities (yes vs. no), rheumatic EIMs (yes vs. no), and presence of a familial history of colorectal colon cancer (yes vs. no).

Statistical analysis
Data are expressed as mean ± standard deviation or median (range) for continuous variables and as frequency (proportion) for qualitative variables.Statistical analysis was performed using the Poisson regression model to obtain the raw and adjusted frequency ratios 15 .For patients with UC, four models were constructed based on four independent dependent variables: history of colectomy, colorectal cancer, hospitalization, and use of biologics or small molecules.The initial regression model was theoretically conceived as having as independent variables such as extensive colitis, proctitis, age at diagnosis, EIMs, comorbidities, BMI, smoking status, sex, and familial history of colon cancer.However, BMI was removed because of numerical insufficiency.Four models were constructed for patients with CD, based on four separately dependent variables: previous surgery, presence of complications (as described above), use of biologics, and hospitalization.As initial independent variables, the models considered the behavior of the disease (B2-stricturing; B3-penetrating), ileal location, colonic location, age at diagnosis, EIMs, comorbidities, smoking status, and sex.For all models, in assessing the contribution of each independent variable to the respective regression model, the module of the percentage difference was calculated for each exponentiated Beta of the regression (frequency ratio) compared with the value 1 as a reference.Patients with IBD-U were excluded because of the small sample size.A value of 5% was set as the minimum permanence value of the variable in the model.The goodness of fit of the models was assessed using the Akaike Information Criterion 16 and residual analysis.To assess the assumption of non-overdispersion, the ratio of the deviance residuals to their degrees of freedom 17 was used.Because the sampling plan was nonprobabilistic, P-values or confidence intervals were not calculated owing to the lack of stability in the standard error estimates 18,19 .The global variance inflation factor (GVIF) was used to evaluate the presence of collinearity, assuming a GVIF < 5% with its absence 20 .Finally, the profile-predictive models were calculated from the final Poisson model equation for models with at least one remaining independent variable.Analysis was performed using the statistical package R (version 4.2.2) 21 in Linux Mint version 21.The statistical review of the study was performed by biomedical statistician.

Ethical considerations
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.This study was approved by the Local Research Ethics Committee, (CAAE: 71343417.2.1001.5629)and by all respective boards from the participating centers (listed in the Declarations).All participants received explanations about the study aims and expected results, having been enrolled in the study only after signing the informed consent term.

Results
A total of 1179 patients were included: 600 (51%) with UC, 568 (48%) with CD, and 11 (0.9%) with IBD-U.Table 1 shows the clinical and demographic characteristics of the patients with IBD.A total of 108 patients were excluded due to inconsistence data.
Regarding the initial symptoms, 42% presented with diarrhea, 38% with abdominal pain, and 20% with weight loss.The age at IBD symptom onset ranged from 1 to 87 years (32.3 ± 14.4) and the time from the beginning of the symptoms to diagnosis was 3.85 years.According to the Montreal Classification of CD, most patients were diagnosed between 17 and 40 years of age (63.1%), the main location was ileocolonic region (41.2%), and the main behavior was non-stricturing and non-penetrating (32%).Among patients with UC, 46.3% presented with extensive colitis, 30% had left-sided colitis, and 23.7% had proctitis.

Clinical factors associated with severity in patients with CD
Table 3 shows the adjusted frequency ratio (FR) for disease severity in patients with CD.The presence of rheumatic EIMs was associated with the use of biological therapy (FR > 1.1).Patients' presenting behavior, according to the Montreal classification, of B2 and B3; colonic location; and presence of rheumatic EIMs demonstrated a higher risk of disease complications (FR > 1.1).Frequency ratios indicated the absence of a contribution of the variables to the model for hospitalization and surgery.
From the final adjusted model, it was possible to obtain the association between the different profiles of patients with CD and the need for biologics.Thus, the most associated patient profile had stricturing disease (B2 behavior), young age at diagnosis, presence of rheumatic EIMs, male sex, and no smoking/no history of smoking status, with an expected average value of 50.24% of the need for biological therapy.The profile most associated with the presence of complications was stricturing or penetrating behavior, not having an ileal location, having a colonic location, being young, having a rheumatic EIM, not having a comorbidity, not smoking, and being male.For such a profile, there is an expected 56% chance of disease complications.

Clinical factors associated with severity in patients with UC
Table 4 shows the adjusted FR for disease severity with respect to UC.The presence of extensive colitis, rheumatic EIMs, male sex, and a familial history of colorectal cancer were associated with the necessity of biological therapy (FR > 1.1).Hospitalization was not associated with any of the variables studied.Regarding the need for colectomy, the associated variables were the presence of extensive colitis, rheumatic EIMs, comorbidities, and male sex (FR > 1.1).
From the final adjusted model, we suggest that patients with extensive colitis, young age at diagnosis, history of colorectal cancer in the family, presence of rheumatic EIMs, male sex, absence of comorbidities, and

Discussion
This study is the first cohort of patients with IBD from the Brazilian National Registry of Patients conducted by the GEDIIB in Brazil.Epidemiological studies are important knowledge tools for a given population aiming for future interventions.Studies with continuous inclusion of patients in their database can generate interesting results in the short, medium, and long terms and can help society, medical organizations, and the government in planning public policies to serve this specific population.A recent increase in the incidence of IBD has been described in the literature, mostly in Asia 22 and Latin America 23 , although a higher prevalence has also been reported in Africa 24 .In absolute numbers, the burden of IBD in developing countries may be greater than that of the combined burden in Europe and North America 25 , which increases costs for the healthcare system, as in most developing countries.In addition, early diagnosis is still lacking, with patients receiving a diagnosis at a later stage with complications.The number of preventable surgeries, outdated treatments (as 5-ASA for CD and corticosteroid use in maintenance therapy), and complications (such as neoplasia) compromise a patient's quality of life.
In Latin America, there are few epidemiological reports from national databases.Juliao-Banos et al. 26 examined the epidemiological trends of IBD in Colombia using a national database of 33 million adults, encompassing 97.6% of the Colombian population.This study calculated the incidence and prevalence of UC and CD from  www.nature.com/scientificreports/2010 to 2017 and examined the epidemiological trends according to urbanicity, demographics, and region.In addition, the IBD phenotype (Montreal Classification), prevalence of IBD-related surgeries, and types of IBD medications prescribed to adult patients attending a regional IBD clinic were assessed.Using a nationally representative sample and a regional clinic cohort, they found that UC is more common in Colombia and is increasing in urban regions.There were 649 patients with IBD in the clinical cohort: 73.7% with UC and 24.5% with CD.The mean ages at diagnosis of CD and UC were 41.0 years and 39.9 years, respectively.Most patients with UC developed extensive colitis (43%), whereas most patients with CD developed ileocolonic disease.A total of 16.7% of patients with CD had perianal disease.Patients with CD received more biologics than those with UC.The National Registry from GEDIIB aims to quantify the real prevalence of IBD in the Brazilian territory as well as to identify the differences between the geographic regions 27 of the country, given the territorial dimension of the same.Our sample demonstrated a predominance of extensive colitis in UC but had a higher percentage of patients with CD (48%).The mean age at the time of IBD diagnosis was 32 years in Brazil and 40 years in Colombia.The time from the beginning of the symptoms to diagnosis was approximately 3.8 years, almost the same as that found by Nobrega et al. 28 and better than that reported by Fróes et al. 27 , who found a delay of 5 years between the reports of the first symptoms and the diagnosis of CD in Rio de Janeiro.
Long-term population studies may have a lower number of patients in their first publications, and it would be interesting to compare population profiles over time.The ECCO-EpiCom 2011 inception cohort analyzed the differences in disease phenotype, medical therapy, surgery, and hospitalization rates during the first year after diagnosis.A total of 258 patients with CD, 380 with UC, and 71 with unclassified IBD were included.Overall, 178 (25%), 460 (65%), and 71 (10%) patients were diagnosed in Eastern Europe, Western Europe, and Australia, respectively.During the first year after diagnosis, surgery and hospitalization rates were significantly higher in patients with CD in Eastern Europe than in those in Western Europe and Australia.In contrast, significantly more patients with CD were treated with biologics in Western European and Australian centers 29 .Our sample included 1179 cases, although modest, a larger number than in the ECCO-EpiCom 2011 inception and Colombia papers, with patients mostly found in the southeastern region of Brazil.
Our study included a representative sample of five regions of Brazil with retrospective and prospective data.It is important to note that the findings presented here reflect only initial data regarding the country's higher number of patients with IBD, as shown by Quaresma et al. 3 .According to their data, IBD has recently reached a prevalence of more than 60 cases per 100,000 inhabitants in Brazil, depending on the degree of urbanization in the region.Therefore, IBD is no longer considered rare.The results of regional studies also show a southnorth gradient, with a higher incidence in urbanized areas in the South/Southeast region of the country 2,3,5,30 .In the states of São Paulo and Espírito Santo 30,31 , prevalence rates of IBD were 52.6/100,000 and 38.2/100,000, respectively in contrast with 12.8/100,000 in the state of Piaui 32 , localized in the northern part of the country, with lower human development index.
To date, most studies in Brazil have analyzed public health data from University Hospitals or the Unified Health System (DATASUS), which is an open-access population-based health and disease registry that contains information from the national unified health system (SUS) through records of hospital admissions, outpatient procedures, consultations, and IBD-related medication dispensing, searched according to the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10), with codes K50 for CD and K51 for UC 2,5,30,31,33 .However, these studies lacked information on private services, which may have led to underestimation.As supplementary health is responsible for approximately 25% of the country's health coverage 34 , a National Registry can contribute to a more accurate epidemiological profile of IBD in the country.In this first analysis of the National Registry results, almost the same number of patients were followed up at private and public (academic or not) centers, suggesting that a relatively high proportion of patients can afford private health insurance.A National IBD Database with inputs from most gastroenterologists and proctologists

Conclusion
To date, no epidemiological study with public and private data has covered the entire Brazilian territory considering the clinical aspects associated with IBD severity.The results obtained from the ongoing registry will be fundamental for improving the information quality in a country with continental dimensions, such as Brazil.The greater the number of qualified participating researchers from different regions of the country, the greater the representativeness of the data, which may greatly help direct government actions on behalf of patients with IBD.

Table 2 .
Previous and current medical treatment distribution among patients with inflammatory bowel disease (n = 1179).

Table 4 .
Clinical characteristics associated with biological therapy, hospitalization, and colectomy in patients with UC (N = 519).FR adjusted frequency ratio; EIM extraintestinal manifestation; UC ulcerative colitis; CRC colorectal cancer.E.g.Significant values are in [bold].