Clinical predictors for a complicated course of disease in an inception cohort of patients with ulcerative colitis: results from the prospective, observational EPICOL study

EPICOL was a non-interventional, prospective, multicentre, observational study in 311 patients with UC as inception cohort (diagnosis, < 6 months) conducted between March 2015 and April 2020. All patients were naive to IS and biologics. Patients were enrolled consecutively at centres representing different levels of care, including outpatient hospital centres and gastroenterological practices in Germany. The endpoints were frequency of UC-related hospitalisation and therapy with IS or initiation of anti–TNF-α therapy. Six visits were scheduled: one at baseline and one each after 3, 6, 12, 18, and 24 months. Any unscheduled visit to a physician, outpatient visits to a hospital, or hospitalisation related to the UC diagnosis were documented as unscheduled visits. Patient demographics, such as age, sex, height, tobacco use, family members with UC, and medical history (e.g. date of first UC symptoms, date of UC diagnosis, previous and current use of non-immunosuppressive UC medication) were collected retrospectively for the status of diagnosis and documented at the baseline visit. In addition, disease symptoms and UC-related extraintestinal manifestations (EIMs), such as pyoderma gangrenosum, erythema nodosum, uveitis/iritis, arthralgia/arthritis, and ankylosing spondylitis, were assessed at every visit and collected retrospectively for the status of diagnosis. Weight and weight loss, localisation of disease, current surgical status, and UC medication or changes in medication (both IS/biologic and non-IS medication) were documented for every visit. Laboratory parameters (anaemia, platelets, faecal calprotectin, C-reactive protein, colonoscopy results, and disease activity according to the simple clinical colitis activity index (SCCAI)) were determined for every visit, if available. All above-mentioned parameters except patient demographics were assessed again at every visit after the baseline visit. The time interval until the need for IS and/or anti–TNF-α therapy and/or UC-related hospitalisation was documented.

EPICOL is a prospective study with the same design as the recently published EPIC (Early Predictive Parameters of Immunosuppressive Therapy in Crohn’s Disease) study, in which we investigated clinical parameters for their predictive power regarding the disease course in patients with recently diagnosed Crohn’s disease (CD) who were naive to treatment with both IS and biologics [24]. In that prospective cohort, we found that a complicated disease course, defined as the need for IS and/or anti–TNF-α treatment and/or CD-related hospitalisation, was significantly associated with various baseline criteria, namely, age at diagnosis < 40 years, anaemia, and treatment with systemic corticosteroids at first flare. Based on these parameters, a risk model in CD was developed that predicted a complicated disease course in our cohort with an accuracy of 87.2%.

In total, 311 patients with a confirmed diagnosis of active UC (except proctitis) no sooner than 6 months before the baseline visit who were ≥ 18 years of age, naive to treatment with conventional IS (thiopurines, calcineurin inhibitors, or equivalent therapy) and biologics (adalimumab, golimumab, infliximab, vedolizumab) at the baseline visit were enrolled consecutively. Pregnant women or patients with previous UC-related surgery were excluded from the study. Note that the use of tofacitinib and ustekinumab for UC was not approved in Europe until July 2018 and September 2019, respectively.

To analyse predictive factors for the need of IS or hospitalisation within 24 months, uni- and multivariate Cox regression analyses were performed. Predictive factors considered overall were sex and clinical response to corticosteroid therapy; predictive factors at diagnosis were age and anaemia; at baseline visit were disease severity (SCCAI), Mayo endoscopic subscore of 3 (severe), and thrombocytosis; and at diagnosis or baseline visit were therapy with systemic corticosteroids, EIM, and smoking status. For the multivariate complete Cox regression model, the factors were examined with regard to co-linearity and interaction. Reduced Cox regression model was established with all factors, with p < 0.05 within univariate or multivariate complete Cox regression models. Parameters that were significant with p < 0.05 in this reduced model were then selected on a stepwise basis to obtain an optimised model. Based on the optimised model, a predictive risk model was developed to determine an individual patient’s probability of experiencing a complicated disease course at 6, 12, or 24 months.

In total, an inception cohort of 311 patients with a recent diagnosis of UC (median time since diagnosis, 1.9 months) was enrolled in the study. Four patients who did not meet inclusion criteria were excluded (“confirmed diagnosis of active UC within the last 6 months” (n = 2) and “immunosuppressive-naïve” (n = 2)), and the final analysis population consisted of 307 patients with UC (Fig. 1). Baseline characteristics of the analysis population are shown in Table 1. The mean age was 38.5 years, and 44.6% (n = 137) were female.

Approximately 50% of all participants had pancolitis; 25% were diagnosed with left-sided colitis and 25% with proctosigmoiditis (Table 2). Regarding the disease history at diagnosis, extraintestinal manifestations were reported for roughly 7% of all participants, of which the most common were arthralgia/arthritis (5.9%). About one fifth of all patients had a documented history of anaemia (n = 58), and weight loss was reported for 93 (30.3%) participants. No history of tobacco use was reported for 184 (59.9%) patients, with 46 (15.0%) and 76 (24.8%) being current and former tobacco users, respectively.

Figure 1 depicts the patient disposition and the proportion of patients with need for IS and/or anti–TNF-α therapy and/or hospitalisation. Of the 307 patients in the analysis population, 209 (68.1%) were neither hospitalised nor received immunosuppressive therapy and were thus classified as having an uncomplicated disease course (group 1). In contrast, 98 (31.9%) patients were classified as having a complicated disease course, requiring hospitalisation and/or IS and/or anti–TNF-α therapy (group 2). Within group 2, 56 (57.1%) were treated with anti–TNF-α therapy either with or without IS, and 42/98 (42.9%) received immunosuppressive monotherapy with azathioprine.

When assessing prior medication, we found that 264 (86.0%) patients had been exposed to non-immunosuppressive therapy, the most common of which was 5-ASA (Table 3). Topical 5-ASA use was reported by 110 (41.7%) patients, and 226 (85.6%) participants reported receiving oral 5-ASA formulations. Prior use of oral budesonide and systemic corticosteroids was documented for 49 (18.6%) and 108 (40.9%) patients, respectively.

At baseline, the mean ± SD disease activity score for the entire cohort as assessed through SCCAI was 3.83 ± 3.16 (Table 4). At baseline, 195 patients (63.5%) were in clinical remission (defined as SCCAI score < 5). The average SCCAI score decreased during the 24-month study period, reaching 1.86 ± 2.4 after 12 months and 1.30 ± 2.10 after 24 months. At the end of the study period, all but 11 patients (n = 190) were in clinical remission.

Examining potential differences of baseline characteristics, we found that patients with a complicated disease course were younger (mean age, 35.7 vs 39.8 years, p = 0.02 Mann–Whitney U test), and more frequently reported weight reduction (45.9% vs 23.0%, p < 0.001, chi-square test) at the time of diagnosis than patients with an uncomplicated disease course. Moreover, they had higher rate of severe Mayo endoscopic scores at baseline (p = 0.033; Supplementary Table S1). We performed a logistic regression analysis to identify baseline parameters that are predictive of a complicated disease course. A list of all examined prognostic factors is provided in Table 4. For this analysis, we relied on the measurements performed at the baseline visit, except for corticosteroid use, which here refers to not only corticosteroid use at baseline but also to previous UC-related corticosteroid use by the patient.

In a first step, all demographic and anamnestic parameters observed in ≥ 5% of patients were analysed with regard to a potential predictive value in both uni- and multivariate regression models (Table 5). All parameters with p < 0.05 in either the uni- or multivariate model (age, male sex, use of corticosteroids, anaemia, and severe Mayo endoscopic subscore of 3 points) were then included in a second multivariate model (Table 6). Parameters that were significant at p < 0.05 in this reduced model were then stepwise selected to be included in an optimised mode. In this final model, prior use of corticosteroids and anaemia were associated with a 2.3- and 1.9-fold increased risk for subsequent need for immunosuppressive therapy and/or hospitalisation, respectively.

Based on the two parameters (therapy with systemic corticosteroids and anaemia), a predictive risk model was developed to determine the individual patient’s likelihood of experiencing a complicated disease course at 6, 12, or 24 months (Table 7).