These guidelines are intended to apply to patients with moderate-to-severe UC disease activity. This is conventionally defined based on the severity of rectal bleeding and diarrhea. According to the PRO2 patient-reported disease activity scale, a stool frequency (SF) score ≥2 and rectal bleeding (RB) score ≥2 suggests moderate-to-severe UC disease activity⁵. Endoscopically, moderate-to-severe UC is indicated by the presence of diffuse erythema, friability, erosions (Mayo endoscopic subscore 2) or spontaneous bleeding or ulcerations (Mayo endoscopic subscore 3). The objective of this guideline is to provide guidance for the pharmacological management of moderate-to-severe UC in outpatients. In addition to patients with moderate-to-severe symptoms, these recommendations are also intended to apply to patients with mildly active symptoms but prognostic signs that predict adverse disease course including high burden of inflammation with severe endoscopic disease activity, corticosteroid dependence or experience significant impact of disease on quality of life. These guidelines also apply to those with moderate-to-severe proctitis. The recommendations in these guidelines do not apply to hospitalized patients with acute severe ulcerative colitis.

The AGA has developed these guidelines as Living Guidelines given rapid evolution in the field⁶. A living guideline is defined as one which allows for optimization of guidelines during the development process with updating of individual recommendations based on the availability of new evidence. Recommendations will only be made for treatments that have received regulatory approval for use in the United States or Europe.

The target audience of these guidelines includes gastroenterologists, advanced practice providers (nurse practitioners or physician assistants), primary care providers, patients, and policy makers. These guidelines are meant to be broad recommendations for management of patients with moderate-to-severe UC and are not intended to address the intricacies of individual patients. Provider experience and patient values and preferences can inform treating providers and patients to reasonably choose alternative treatment options.

This document represents official recommendations from the AGA. It was developed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework for therapeutic strategies and adheres to best practices in guideline development, per the direction provided by the National Academy of Medicine⁷. The development of this document is fully supported by the AGA Institute.

Members of the guideline panel were selected based on clinical and methodological expertise and experience, and after review of all conflicts of interest in a comprehensive vetting process. The multidisciplinary guideline panel included gastroenterologists with expertise in inflammatory bowel disease (IBD) and guideline methodologists. Panel members disclosed all conflicts of interest, which were defined and categorized per AGA policies and the National Academy of Medicine and Guidelines International Network standards^{7, 8}. No guideline panel member was excused from participation in the process owing to disqualifying conflict. A full list of conflicts can be accessed at AGA’s National Office in Bethesda, MD.

Using the PICO format, which frames a clinical question by defining a specific Population (P), Intervention (I), Comparator (C), and Outcomes (O), the team finalized 12 questions to be addressed (Table 1). The AGA Governing Board approved the final set of questions and statements in September 2023. Consistent with AGA’s prior guidelines on the pharmacological management of moderate-to-severe UC⁴, induction and maintenance of clinical remission were considered critical outcomes for decision-making, whereas achieving endoscopic remission, endoscopic improvement, corticosteroid-free remission, serious adverse events, serious infections and treatment tolerability (drug discontinuation due to adverse events) were considered important outcomes. Clinical remission was most commonly measured using the Mayo Clinic score (MCS), an index with scores ranging from 0–12, based on measures of stool frequency, rectal bleeding, physician global assessment, along with endoscopic disease activity⁹. Scores of 6–12 correspond to moderate-to-severe disease activity, whereas clinical remission is most consistently defined as MCS<3, with no individual subscore >1. In earlier trials, alternative cut-offs of MCS-defined remission and alternative disease activity indices such as Powell-Tuck index, Baron endoscopy score, and others were used¹⁰. In these trials, if clinical and endoscopic outcomes were reported separately, then data on clinical remission was used for analysis. While RCTs also variably report clinical response outcomes, attaining clinical remission is the primary treatment target in moderate-to-severe UC and thus was the primary outcome used to grade evidence and inform absolute and comparative treatment efficacy.

To provide a synthesis of the risks and benefits of different interventions, and to calculate absolute effect estimates, the panel relied on pooled clinical remission rates on placebo. In randomized controlled trials (RCT) with advanced therapies, the rate of induction of clinical remission with placebo was set at 10%, and maintenance of clinical remission was set at 15%. In trials of thiopurines which reported steroid-free remission as an outcome, pooled rates across placebo arms were used. For comparisons against no treatment, the guideline panel set a clinically meaningful difference (CMD) threshold of 10%, based on consensus. If the effect size was below this CMD threshold, then benefit was deemed to be trivial. For comparisons between two active therapies, the guideline panel set a CMD threshold of 5%, based on consensus i.e. we considered the difference between an active agent vs. comparator as ‘important’ if the absolute risk difference of achieving remission crossed the CMD threshold of >50 per 1000 patients treated (5%).

A comprehensive search of Ovid MEDLINE, Embase, and Wiley Cochrane Library, using a combination of controlled vocabulary terms and relevant keywords (Supplementary Table 1), from inception to November 21, 2023, was conducted by an experienced medical librarian, with input from the guideline methodologist. The search was updated on September 1, 2024. In addition, we reviewed references of previous guidelines and consensus statements, conference proceedings and press releases on novel advanced therapies. Content experts provided insights into ongoing studies. All searches were limited to human subjects and English language. For evidence synthesis, RCTs conducted in adults with moderate-to-severe UC (corresponding to relevant PICOs) were included. If RCT-level evidence was not available for specific PICOs, then observational studies were considered to inform evidence. Due to the relatively recent approval of several of the medications under consideration, there is a paucity of real-world data regarding their use, both for effectiveness and safety outcomes. The minimum trial duration for induction and maintenance therapy was 4 weeks and 24 weeks, respectively. Efficacy trials exclusively in patients with Crohn’s disease (CD) were excluded (except for data on combination therapy and treatment de-escalation); if a trial included both patients with UC and CD, it was included only if results were stratified by disease type or if >70% participants had UC. Since safety outcomes are not well informed by RCTs, representative large cohort studies and high-quality systematic reviews/meta-analyses were used to inform risk of serious infections and malignancy with different therapies. In addition, studies on issues of racial, ethnic, and social disparities and issues of general health equity pertinent to the topic were identified. Data abstraction was conducted in duplicate, independently, by two sets of investigators, with disagreements or questions of accuracy resolved by discussion and consensus.

For trials of induction and maintenance therapy, outcomes were abstracted and reported as induction of clinical remission (in patients with active disease), and maintenance of remission (in patients with quiescent disease at trial entry), respectively. All analyses were conducted using true intention-to-treat analysis; patients lost to follow-up or excluded from analysis for other reasons were deemed to be treatment failures. Pooled relative risk (RR) or odds ratios (OR) and 95% confidence intervals (CI), were calculated using the Mantel-Haenszel fixed-effects model (in the absence of conceptual heterogeneity and if <5 studies) or the DerSimonian-Liard random-effects model¹¹. First, pooled relative risk estimates were derived from pooled clinical trials for each individual therapy; these risk estimates are influenced both by efficacy of individual therapies but also the placebo rates in each of these trials. These relative risk estimates were then applied to a standardized placebo rate that represented the pooled placebo rate across all Phase 2 and Phase 3 RCTs in moderate-to-severe UC. The absolute risk difference derived from this was then used to inform strength of evidence for efficacy for each treatment. Use of a pooled standardized placebo response rates minimizes the effect of variable placebo rates across each trial while still being informed by the efficacy of each treatment from the parent RCTs and enhances generalizability. This choice of pooled placebo rates is consistent with prior AGA guidelines for management of ulcerative colitis and Crohn’s disease and recommendations from GRADE. Statistical heterogeneity was assessed using the I² statistic¹². Small study effects were examined using funnel plot symmetry and Egger’s regression test, though it is important to recognize that these tests are unreliable when the number of studies is <10¹³. Direct comparisons were performed using Comprehensive Meta-Analysis, v2.0. Due to a paucity of head-to-head trials of active agents, to inform comparative efficacy of different pharmacologic interventions, we performed network meta-analysis (NMA) using the frequentist approach, with the statistical package “netmeta” (version 9.0, https://cran.r-project.org/web/packages/netmeta/index.html) in R (version 4.0.2). Details of the NMA are reported in the accompanying 2024 AGA Evidence Synthesis document on comparative efficacy of different advanced therapies for management of moderate-to-severe UC that has been co-published in the Journal.

The quality of evidence was judged using the GRADE framework¹⁴. Briefly, using this approach, evidence from RCTs starts at high quality, and evidence from observational studies starts at low quality evidence (Supplementary Table 2). This evidence can be further rated down for risk of bias in the evidence, indirectness, inconsistency, imprecision, and publication bias. In selected cases, particularly for observational studies, evidence may be rated up if a large treatment effect is observed, if there is a dose-response relationship or if all plausible confounding and bias would reduce a demonstrated effect or suggest a spurious effect if no effect was observed. Evidence profiles were developed for each intervention using the GRADEpro Guideline Development Tool (https://gradepro.org).

For questions of comparative efficacy of different pharmacological interventions, we used GRADE approach for NMA. Details are reported in the accompanying 2024 AGA Evidence Synthesis document that has been co-published in the Journal. In this evaluation, we considered the difference between an active agent vs. comparator as ‘important’ if the absolute risk difference of achieving remission crossed the CMD threshold of >50 per 1000 patients treated (5%), and ‘trivial’ if the absolute risk difference was between 0 to 50 per 1000 patients treated. In utilizing NMA for evidence synthesis, we relied on direct evidence when it was available from head-to-head comparisons and provided at least moderate certainty evidence. If there were no direct comparisons between two interventions or if the evidence from direct comparisons was very low or low certainty evidence, then effect estimates from the NMA were used. O

Based on the GRADE Evidence-to-Decision framework, the guideline panel weighed the magnitude of and balance between the benefit and harms of interventions, patients’ values and preferences, and the domains of feasibility, acceptability, and resource requirements and the impact on health equity¹⁵. The panel reached a consensus for all guideline statements. The certainty of evidence and the strength of recommendation are provided for each clinical question. Based on GRADE methodology, we label recommendations as “strong” or “conditional.” The phrase “we recommend” indicates strong recommendations and the phrase “we suggest” indicates conditional recommendations and provide the suggested interpretation of strong and weak recommendations for patients, clinicians, and health care policy makers (Table 2). In addition, the panel provided broad overarching, as well as recommendation-specific implementation considerations to provide context and facilitate real-world use and adoption of these recommendations, based on evidence and their clinical experience and practice.

This guideline was submitted for public comment and external peer review and was approved by the AGA Governing Board. The accompanying 2024 AGA Evidence Synthesis document focusing on comparative efficacy of different advanced therapies underwent conventional peer review.

The guideline panel rated the safety of pharmacological therapies as an important but not critical outcome for decision-making. It is important to note that clinical trials are selective in enrollment, and often have short follow-up. Data from these trials are often not able to adequately assess the safety of different therapies. Hence, we reviewed large cohort studies and published systematic reviews and meta-analyses to understand comparative safety of different advanced therapies in patients with UC. In a systematic review and meta-analysis of 20 head-to-head studies comparing risk of infections between different advanced therapies for treatment of IBD, Solitano and colleagues observed that vedolizumab was associated with a 32% lower risk of serious infections compared with tumor necrosis factor (TNF) antagonists in patients with UC with minimal heterogeneity¹⁶. In patients with CD, ustekinumab was associated with 51% lower risk of serious infections compared with TNF antagonists, and 60% lower risk compared with vedolizumab. There have been limited comparative safety studies of new small molecule drugs including Janus kinase (JAK) inhibitors and sphingosine-1-phosphate (S1P) receptor modulators in patients with IBD. In a US administrative claims-based study, Cheng et al. did not observe any significant difference in the risk of all- or serious infections between 305 patients with IBD treated with tofacitinib vs. 19,096 patients treated with TNF antagonists¹⁷. In contrast, in the large ORAL surveillance trial in older adults (≥50 years) with rheumatoid arthritis and at least one cardiovascular risk factor, tofacitinib was associated with a higher risk of serious and opportunistic infections compared with TNF antagonists¹⁸. There are limited real-world data on safety of newer advanced therapies like S1P receptor modulators and interleukin (IL)-23 antagonists, particularly in patients with UC. Comparative safety studies of JAK inhibitors with non-TNF antagonist biologics are sparse. Across studies, most consistent risk factors for serious infections are disease-related (e.g., high disease activity, inadequate disease control, and need for corticosteroids and opioids), and individual patient-related (e.g., advanced age, frailty and comorbidities)^{17, 19–24}.

TNF antagonists have also been associated with increased risk of lymphoma and melanoma. In a French population-based study, Lemaitre and colleagues estimated the annual incidence of lymphoma in patients treated with TNF antagonist monotherapy vs. unexposed patients to be 0.41 per 1000 person-years vs. 0.26 per 1000 person-years; after adjusting for potential confounders, the risk of lymphoma was 2.4-times higher in patients treated with TNF antagonist monotherapy²⁵. This risk was comparable to the risk observed in patients treated with thiopurine monotherapy. Patients exposed to combination therapy had six-fold increased risk of lymphoma, as compared to unexposed patients, and 2.3–2.5 times higher risk as compared to patients exposed to monotherapy with either agent. The FDA has issued a black box warning on the increased risk of malignancy with TNF antagonists (https://www.accessdata.fda.gov/drugsatfda_docs/label/2013/103772s5359lbl.pdf). Currently, there is a paucity of population-representative data to inform risk estimates related to malignancy for other classes of advanced therapies.

Overall, the guideline panel felt that while advanced therapies, particularly TNF antagonists, JAK inhibitors and S1P receptor modulators, may be associated with increased risk of serious infections, the magnitude of increased risk is small for most patients with moderate-to-severe UC, and overall balance significantly favored benefits over harms with these agents for most patients. In addition, active disease and ongoing corticosteroid use to control symptoms are important determinants of safety outcomes; consequently the efficacy of treatments in achieving remission, particularly steroid free remission, is important in reducing overall likelihood of infections for patients. Where there are individual patient characteristics pertaining to treatment safety that may influence selection of therapy, they are discussed below.

The data examining the efficacy of advanced therapies against placebo was derived from 38 phase 2 or phase 3 RCTs of approved treatments for moderate-to-severe UC. This comprised 13 trials of TNF antagonists (5 infliximab^26–29, 5 adalimumab^30–33, 3 golimumab^34–36), three trials of anti-integrins (3 vedolizumab^37–39), one trial of ustekinumab⁴⁰, five trials anti-IL23 antibodies (2 mirikizumab^{41, 42}, 1 risankizumab⁴³, 2 guselkumab^44–46), five trials of S1P receptor modulators (2 ozanimod^{47, 48}, 3 etrasimod^{49, 50}), and eight trials of JAK inhibitors (3 upadacitinib^{51, 52}, 3 tofacitinib^{53, 54}, 2 filgotinib⁵⁵). Of the included RCTs, the trials of ustekinumab, filgotinib, upadacitinib, ozanimod, etrasimod, mirikizumab, risankizumab and guselkumab were all new since the previous 2020 guideline evidence synthesis⁵⁶. For infliximab and vedolizumab, new information since the 2020 guideline included efficacy of the subcutaneous formulations of each drug for maintenance of remission. For adalimumab, additional data on efficacy was available in the placebo controlled RCT comparing adalimumab and etrolizumab³⁰. All trials were conducted in patients with moderate-to-severe UC and compared efficacy against placebo. Patient characteristics including severity were broadly comparable across all trials; however, later trials had a larger proportion of patients with multiple biologic failures prior to study entry. These are summarized in greater detail in the accompanying evidence synthesis document. Most trials provided information on both biologic-naïve and biologic-exposed patients except for trials of infliximab and golimumab which included only biologic-naïve patients. Trials of biosimilars (infliximab, adalimumab) or alternate modes of delivery such as subcutaneous injections (infliximab, vedolizumab) were also included when applicable.

Data on the efficacy of each agent vs. placebo for induction and maintenance of clinical remission is shown in eFigures 1–13. The corresponding GRADE evidence profile with certainty of evidence for each agent is shown in Tables 5–9. Overall, upadacitinib was superior to placebo for inducing and maintaining clinical remission with high certainty of evidence. Infliximab, adalimumab, golimumab, vedolizumab, tofacitinib, filgotinib, ustekinumab, ozanimod, etrasimod, mirikizumab, risankizumab and guselkumab were superior to placebo with moderate certainty of evidence.

In order to make recommendations, it is critical to examine the benefits and harms of choosing an intervention over a comparator. Overall, given the multiplicity of treatments available for moderate-to-severe UC and the importance of making informed, evidence-based, and cost-effective choices, the panel established an CMD of 10% over placebo to suggest at least a moderate desirable effect with the intervention. Infliximab, golimumab, vedolizumab, ustekinumab, tofacitinib, ozanimod, etrasimod, risankizumab and guselkumab were deemed to have moderate desirable effect, whereas upadacitinib was deemed to have a large desirable effect (≥ 20%). In contrast, adalimumab, mirikizumab, and filgotinib were deemed to have trivial to small desirable effect over no intervention, since the magnitude of benefit was below the pre-specified CMD. All active interventions were deemed to have trivial undesirable effects relative to no intervention given the low risk of treatment-related serious adverse events with these therapies such as serious infections and malignancy. Importantly, the panel considered potential harms of no intervention to include risks associated with untreated disease which could negatively impact quality of life, functional status and leads to greater need for corticosteroids, and themselves could increase the risk of serious infections and in some instances, malignancy such as colorectal cancer. The GRADE evidence-to-decision judgements for use of all advanced therapies over no intervention is shown in Table 10.

The panel surmised that patient characteristics were broadly similar across the clinical trials although the later clinical trials were more enriched with biologic-refractory patients. The panel recognized the heterogeneity in the response of placebo-treated patients across the trials which could affect the magnitude of relative risk reduction. Consequently, absolute risk differences informing strength of evidence for each agent were made against a standardized placebo response rate that represented an average across trials.

As noted earlier, in the implementation of the guideline recommendations, it is important to factor in patient-related factors to guide selection of therapy. These include assessment of patient risk for immunosuppression related complications including infections or prior malignancy, presence of extra-intestinal manifestations or other disease-complications that may influence therapy, and patient preference for route of administration.

Three biosimilars for infliximab, ten for adalimumab, and one for ustekinumab have been approved for use for moderate-to-severe UC in the United States. In RCTs, switching from parent to biosimilar infliximab was not associated with higher rates of relapse⁵⁷. In observational studies, most patients tolerated a switch to a biosimilar without an increase in loss of response or adverse events^{58, 59}. Thus, patients newly starting, or on established therapy with, infliximab, adalimumab, or ustekinumab, treatment outcomes with originator or biosimilar in most situations may be comparable. There is no increase in the risk of immunogenicity because of this switch; existing drug assays are accurate in measuring biosimilar trough levels with therapeutic thresholds interchangeable with originator drug⁶⁰.

Subcutaneous formulations of infliximab and vedolizumab have been approved as maintenance therapy for patients with moderate-to-severe UC. Phase 3 RCT enrolled patients with moderate to severely active UC to receive open-label infliximab biosimilar (CT-P13) 5mg/kg intravenously at weeks 0, 2, and 6⁶¹. Responders were randomized to receive either CT-P13 120mg subcutaneous (SC) or placebo every 2 weeks for up to week 52. At the end of the trial, the rates of clinical remission were higher with CT-P13 (43.2%) compared to placebo (20.8%). The VISIBLE 1 trial examined the efficacy of SC vedolizumab following 2 intravenous (IV) induction doses to standard IV maintenance therapy with vedolizumab or placebo⁶². At week 52, the rates of remission were similar with SC vedolizumab (46.2%) and IV vedolizumab (42.6%) and higher than placebo (14.3%). Real-world experience suggests high acceptability and comparable effectiveness with switching from intravenous to subcutaneous formulations⁶³. The guideline panel felt that subcutaneous formulations of infliximab and vedolizumab are acceptable alternatives to intravenous maintenance therapy for most patients. Dosing considerations should be factored in for patients with severely active disease, high body mass index and those on dose-escalated regimens.

Emerging data suggest that in some patients, particularly those with severe disease, an extended induction regimen is necessary to improve rates of clinical response particularly with JAK inhibitors. Among patients who did not respond to the 10mg twice daily (BID) dose of tofacitinib at week 8 in the OCTAVE trial, 52% achieved clinical response at week 16 following extended induction with 10mg BID dosing for an additional 8 weeks⁶⁴. Of these, 56.1% maintained clinical remission at 36 months. Nearly half (48%) of patients who failed to respond to the initial 8-week induction regimen with upadacitinib 45mg/day responded to an additional 8 weeks of induction therapy⁶⁵. Over half of these patients maintained clinical response at 1 year. Similarly, an additional three IV induction doses of mirikizumab showed benefit in patients with incomplete response to the first three weeks induction dosing⁶⁶.

Patients, particularly those with severe disease, may also require maintenance therapy at a higher dose. In a study of de-escalation of tofacitinib to the 5mg BID maintenance dose, approximately 29% of patients who were de-escalated required an increase in dose back to 10mg BID with clinical response re-capturable in only 63% of patients⁶⁷. Similarly in the OCTAVE trials, 25% patients who were de-escalated were not able to remain in remission on the lower maintenance dose⁶⁸. Thus, a subset of patients may require being maintained at a higher dose. Since some of the risk of adverse effects associated with tofacitinib, particularly shingles and VTE are greater at the higher dose, it is important to monitor such patients carefully for these events and adopt preventive strategies to minimize their risk.

The positioning of therapies is a critical component of management of moderate-to-severe UC. Recognizing the paucity of head-to-head trials, the panel relied on NMA to inform comparative efficacy of different agents. The accompanying evidence synthesis document summarizes in detail the results of the NMA. Briefly summarizing the results of the NMA for induction of clinical remission in patients with moderate-to-severe UC who are naïve to advanced therapies, several pair-wise comparisons between two active treatments met the a priori threshold for superiority (CMD ≥5%). Infliximab has a possibly important benefit in achieving remission compared with adalimumab, mirikizumab, tofacitinib, and filgotinib with a low certainty of evidence. Golimumab, similarly, has possibly important benefit over adalimumab, filgotinib, and tofacitinib, with a low certainty of evidence. Vedolizumab achieves a possibly important benefit compared with adalimumab (including from the head-to-head VARSITY trial³⁸) and tofacitinib with a low certainty of evidence. Ozanimod demonstrated a possibly important benefit over adalimumab, mirikizumab, tofacitinib, and filgotinib while etrasimod demonstrated a possibly important benefit over filgotinib, with a low certainty of evidence. Risankizumab likely has important benefit compared with filgotinib with moderate certainty evidence and possibly important benefit compared with adalimumab, ustekinumab, mirikizumab and tofacitinib. Guselkumab demonstrated a possibly important benefit over adalimumab, mirikizumab, tofacitinib and filgotinib with low certainty of evidence. Upadacitinib demonstrated a likely important benefit over infliximab, adalimumab, vedolizumab, ustekinumab, mirikizumab, etrasimod, tofacitinib and filgotinib with moderate certainty of evidence, and possibly important benefit over golimumab and ozanimod with a low certainty of evidence. Analyses of endoscopic improvement after induction showed findings broadly consistent with induction of clinical remission. It was difficult to compare the efficacy of treatments for maintenance of clinical remission through a NMA approach because of heterogeneity in the trial design. The only head-to-head trial was the VARSITY trial where 769 patients with moderate-to-severe UC were randomized to receive either vedolizumab (n=383) or adalimumab (n=386)³⁸. At week 52, a higher rate of clinical remission was observed in the vedolizumab treated patients (31.3%) compared to adalimumab (22.5%). Corticosteroid-free remission rates, however, were higher with adalimumab (21.8%) compared to vedolizumab (12.6%). On NMA of treat-straight-through trials, etrasimod demonstrated likely important benefit over infliximab and possibly important benefit over adalimumab.

No significant differences in the risk of infections and serious adverse events between different agents in previous network meta-analyses of clinical trials¹⁶. In observational studies, vedolizumab was associated with a lower risk of serious infections compared with TNF antagonists in patients with UC^{69, 70}; there was paucity of evidence for comparative safety of other agents. Thus, they may be preferred among agents of similar efficacy in patients particularly vulnerable to infectious complications such as the older frail adult or those with recent malignancy, excluding non-melanoma skin cancers. Specific safety considerations regarding pregnancy and lactation are discussed in the implementation consideration below. Overall, the panel established a CMD of 5% over an active intervention to suggest at least a moderate desirable effect with the intervention relative to the comparator; for most comparisons, this risk difference was between 5–15%. Only upadacitinib was deemed to have a large desirable effect exceeding 20% for comparisons against infliximab, adalimumab, vedolizumab, ustekinumab, mirikizumab, etrasimod, tofacitinib and filgotinib. The panel deemed that even though there may be small differences in the relative risk of adverse events with different medications, the overall magnitude of these undesirable effects with all medications was trivial. Hence, decision-making for recommendations was driven primarily by comparative efficacy, rather than safety, in most instances.

With the availability of multiple drugs within a class for several therapeutic mechanisms, there are emerging data suggesting differences in efficacy even within a therapeutic class. Consequently, the panel made recommendations specific to individual drugs rather than to classes. Rather than relying only on active comparisons showing important benefit over a comparison through NMA, the guideline panel took a more pragmatic and holistic approach to create ‘efficacy buckets’ that grouped together treatments with similar magnitude of treatment benefit. The efficacy buckets were informed by magnitude of absolute risk difference (over placebo) in the Phase 3 RCTs as well as comparative efficacy in the NMA. Treatments were generally considered high efficacy if they demonstrated a CMD ≥5% in direct or network head-to-head comparison (if no direct evidence available), a p-score in the NMA of 0.49 or higher, and an absolute benefit of ≥15% over placebo in phase 3 RCT among biologic-naïve patients. Recognizing several medications have only recently been approved and there is paucity of real-world evidence on their absolute and comparative effectiveness, the panel took a more conservative approach on creating efficacy buckets. Applying the above criteria, the guideline panel classified infliximab, vedolizumab, ozanimod, etrasimod, upadacitinib, risankizumab and guselkumab as high efficacy medications while golimumab, ustekinumab, tofacitinib, filgotinib and mirikizumab were labeled moderate efficacy medications, and adalimumab was rated as having lower efficacy. The panel debated the relative weight of efficacy during induction and maintenance phases to inform comparative effectiveness and relative positioning. Given the recognized adverse impact of corticosteroid therapy and the importance of achieving steroid free remission early as a treatment endpoint, the panel prioritized the relative efficacy of treatments in achieving clinical remission at the end of the induction as primary determinant of relative positioning of treatments. However, patients and providers may reasonably select medications based on strength of maintenance efficacy, particularly in the setting of other patient- or disease-factors (such as comorbidity) that may influence treatment choice.

The panel recognized the varying recommendations for use of JAK inhibitors as first-line treatment in different regions of the world. In the United States, the FDA label recommends use of JAK inhibitors in patients with prior failure or intolerance to TNF antagonists. In Europe, the EMA recommends cautious use of JAK inhibitors as a first line agent in patients at risk for adverse cardiovascular outcomes including age ≥ 65 years, current or previous long-term smokers, history of cardiovascular disease (such as heart attack or stroke), and patients with a history of cancer. Restriction for use of JAK inhibitors first line is largely informed by data from the ORAL Surveillance study¹⁸. This study compared the safety of tofacitinib (5mg or 10mg BID) to TNF-antagonist therapy among patients with rheumatoid arthritis aged 50 years or older who also had cardiovascular risk factors (such as prior cardiovascular disease or smoking cigarette smoker [current or past], hypertension, high-density lipoprotein cholesterol level of <40 mg per deciliter, diabetes mellitus, family history of premature coronary heart disease, extraarticular rheumatoid arthritis, or history of coronary artery disease). In this cohort, over a median follow up of 4 years, patients using tofacitinib, particularly at the 10mg BID dose, had a higher risk of major adverse cardiovascular events (MACE) including venous thromboembolism and cancer. In regions where JAK inhibitors may be used as first line therapy in biologic-naïve patients, upadacitinib can be considered a high-efficacy medication while tofacitinib and filgotinib are considered moderate-efficacy medications. In patients at high risk of MACE, JAK inhibitors should be used cautiously.

For patients who desire an oral route of administration, an S1P receptor modulator or JAK inhibitor may be a preferred therapeutic agent. Large prospective registries have demonstrated that maternal use of TNF antagonists or other biologics during pregnancy is not associated with a significant increase in risk of adverse pregnancy or early childhood outcomes^{71, 72}. While there is limited data on newer anti-IL23 inhibitors (mirikizumab, risankizumab, guselkumab), it can be reasonably expected that their safety profile during pregnancy will be similar to ustekinumab. In contrast to above, there is a paucity of data on small molecule treatments including JAK inhibitors and S1P receptor modulators with animal data suggesting potential adverse pregnancy effects at doses much higher than used for treatment of IBD⁷³. Thus, in women of childbearing age actively contemplating pregnancy, we recommend avoiding JAK inhibitors and S1P receptor modulators and selecting an alternate therapeutic option when possible.

The relative positioning of different therapies was informed primarily by comparative efficacy in inducing clinical remission which was defined by the panel a priori as a critical outcome of interest. The panel recognized that other patient-important endpoints including achieving corticosteroid-free remission, maintenance of clinical remission, avoiding surgery and hospitalization as well as objective outcomes such endoscopic and histologic healing are important treatment goals for moderate-to-severe UC. The heterogeneity in trial designs (responder re-randomization or treat-straight-through) prevented robust comparisons for treatments for longer-term endpoints. Where data were available, relative efficacy for maintenance endpoints was broadly consistent with induction data. Similarly, data on achievement of endoscopic improvement were also consistent with clinical remission endpoints. There was lack of systematic reporting of other endpoints particularly for older clinical trials that precluded using such data to inform relative positioning.

The body of evidence for comparative effectiveness of individual therapy in patients exposed to advanced therapies is summarized in the accompanying NMA. There was a single head-to-head randomized trial (VARSITY) comprising a small number (21%) of previously biologic-exposed patients comparing vedolizumab and adalimumab in moderate-to-severe UC³⁸. This trial noted no difference between the two agents in maintaining clinical remission (20.3% vs 16.0%) in this subpopulation. Thus, most of the evidence for relative positioning was informed by the NMA as well as direct evidence from the Phase 2 and Phase 3 RCTs. In this synthesis, tofacitinib, filgotinib, upadacitinib, ustekinumab, mirikizumab, risankizumab and guselkumab demonstrated likely important benefit in achieving clinical remission compared with no treatment with a moderate certainty of evidence. Amongst active comparisons, upadacitinib likely has important benefit over adalimumab, vedolizumab, filgotinib, etrasimod, mirikizumab, risankizumab and guselkumab with a moderate certainty of evidence, and possibly important benefit over ozanimod. Tofacitinib also demonstrated likely important benefit over adalimumab, vedolizumab, and etrasimod, and possible important benefit over ozanimod and mirikizumab with a low certainty of evidence. Ustekinumab demonstrated a likely important benefit over adalimumab, vedolizumab, ozanimod, etrasimod, and mirikizumab with a moderate certainty of evidence. Risankizumab was associated with possibly important benefit over adalimumab, vedolizumab and etrasimod with a low certainty of evidence. Guselkumab was associated with likely important benefit over adalimumab with moderate certainty of evidence, and possibly important benefit over vedolizumab and etrasimod with low certainty of evidence. Notably, there was no RCT for infliximab or golimumab in patients with prior exposure to biologics.

The comparative safety of individual therapeutic agents is discussed above in Questions 1 and 2 and “Safety of Pharmacological Therapies” section. Overall, the panel established a clinically meaningful difference (CMD) ≥5% over an active intervention to suggest at least a moderate desirable effect with the intervention relative to the comparator; for most comparisons, this risk difference was between 5–10%. Tofacitinib, upadacitinib and ustekinumab demonstrated large desirable effect exceeding 30% for relevant comparisons. The panel deemed that even though there may be small differences in the risk of adverse events with different medications, the overall magnitude of these undesirable effects with all medications was trivial. Importantly, while observational studies suggest lower risk of serious infection with vedolizumab compared with TNF antagonists, inadequate disease control is also associated with increased risk of infections²³. In these instances, any potential safety benefit of more targeted advanced therapies must be balanced against differences in treatment efficacy. This is particularly relevant for biologic-exposed patients where safer drugs (such as vedolizumab) were rated lower in efficacy than other, less-targeted mechanisms.

The panel recognized differences in treatment efficacy between drugs within the same therapeutic class. Consequently, recommendations were made specifically for drugs rather than by therapy class. As in the biologic-naïve population, rather than relying only on active comparisons showing important benefits over a comparator in an NMA, the guideline panel took a more pragmatic and holistic approach to create ‘efficacy buckets’ that grouped together treatments with similar magnitude of treatment benefit. The efficacy buckets were informed by magnitude of absolute risk difference (over placebo) in the Phase 3 RCTs as well as comparative efficacy in the NMA. Treatments were generally considered high-efficacy if they demonstrated a CMD of ≥5% or greater in direct or network head-to-head comparisons (if no direct evidence available), a p-score in the NMA of 0.49 or higher, and an absolute benefit of >10% over placebo in Phase 3 RCTs among biologic-exposed patients. Based on these criteria, tofacitinib, upadacitinib, and ustekinumab were considered higher-efficacy medications, filgotinib, mirikizumab, risankizumab and guselkumab were considered intermediate-efficacy medications while adalimumab, vedolizumab, ozanimod, and etrasimod were considered lower-efficacy medications in biologic-exposed patients with moderate-to-severe UC.

The panel recognized several considerations for the interpretation of data in the biologic-exposed patients. Of the studies that examined efficacy in biologic-exposed patients, >90% patients had exposure to TNF antagonists. Up to half the patients in the later treatment trials were also exposed to vedolizumab or other therapeutic mechanisms. Given this, no recommendations could be made for patients who had specifically failed only prior non-TNF advanced therapies. Recommendations were also made broadly for biologic-exposed patients, but the number of prior biologic failures may impact treatment efficacy. In the Phase 3 randomized trial of ozanimod against placebo, ozanimod was significantly more effective than placebo in achieving clinical remission with rates of 23% (vs 7%) and 17% (vs. 8%) in biologic-naïve and single biologic-exposed patients. In contrast, in patients who had been on ≥2 biologics previously, ozanimod was no more effective than placebo (4% vs. 3%) in inducing clinical remission. However, the efficacy of ozanimod in achieving clinical remission at week 52 was similar for the biologic-naïve, 1 biologic, and ≥ 2 biologic exposed patients, with all differences being greater than placebo⁷⁴. Thus, in patients with multiple prior biologic-failures, a longer duration of treatment may be required for clinical benefit for treatments in the lower-efficacy category. As noted above, there are limited data on use of JAK inhibitors or S1P receptor modulators during pregnancy. Thus, other agents should be preferred over them where possible in women contemplating pregnancy. These decisions should be made on an individual case-by-case basis recognizing the impact of uncontrolled disease on fertility and pregnancy outcomes and using personalized risk-benefit thresholds.

The differences in trial designs between the various agents precluded our ability to compare efficacy in maintaining clinical remission or endoscopic improvement to inform positioning. There are no RCTs that examined infliximab or golimumab in patients with prior biologic exposure. However, observational studies have demonstrated effectiveness of both these agents in this setting and it is reasonable to consider them in patients with prior TNF antagonist exposure, particularly those who discontinued prior therapy due to intolerance and/or immunogenicity^{75, 76}.

Since the last guideline published in 2020, we identified one new RCT comparing mercaptopurine to placebo for achieving corticosteroid-free remission at 52 weeks⁷⁷. In this trial, patients with active UC despite 5-ASA were randomized 1:1 to therapeutic drug monitoring-guided mercaptopurine vs. placebo for 52 weeks; all patients received corticosteroids for the first 8 weeks. The primary endpoint of corticosteroid-free clinical remission was achieved by 14/29 patients treated with mercaptopurine compared to 3/30 patients treated with placebo. Including this trial, we identified four trials comparing thiopurines to placebo and two trials comparing thiopurines to 5-ASA for inducing corticosteroid free remission. In four out of six trials, patients were considered corticosteroid-dependent, unable to taper corticosteroids below 10–20mg/day without relapsing. In contrast to more recent clinical trials, different disease activity indices were used in these studies, and the outcome of corticosteroid-free remission was assessed at variable intervals from 4 weeks to 52 weeks. In patients with active disease, patients were started simultaneously on thiopurines and corticosteroids, and it was unclear whether remission was induced by corticosteroids or thiopurines or the combination of both. Thiopurines were associated with a higher rate of achieving corticosteroid-free clinical remission compared to placebo or 5-ASA (RR, 1.41; 95% CI, 0.91–2.18) (eFigure 14). However, the overall quality of evidence was deemed very low, due to serious risk of bias, imprecision, and indirectness (outcome definition and assessment) (Table 11). On limiting analysis to studies where outcome was assessed 26 weeks or greater, thiopurines were associated with a higher rate of corticosteroid-free clinical remission compared to placebo or 5-ASA (RR, 2.62; 95% CI, 0.99–6.96).

Since the last guideline published in 2020, we did not identify any new RCTs examining the efficacy of thiopurines in preventing relapse in patients with quiescent UC⁵⁶. For maintenance of remission, we identified four trials comparing thiopurines to placebo and three trials comparing thiopurines to 5-ASA. Maintenance of remission was defined as prevention of relapse following corticosteroid-induced remission (5 trials) or as the ability to maintain a corticosteroid-free remission in patients on long-standing thiopurine therapy (2 trials), evaluated between 6–18 months. On meta-analysis, thiopurines were more effective than placebo or 5-ASA for decreasing risk of disease relapse (RR, 0.61; 95% CI, 0.49–0.77) among patients with inactive UC (in remission) (eFigure 14). There was low certainty of evidence due to risk of bias and imprecision (Table 11).

Since the last guideline published in 2020, we did not identify any new RCTs examining the efficacy of methotrexate for induction or maintenance of remission in patients with UC⁵⁶. Two trials compared methotrexate to placebo and one trial compared methotrexate to 5-ASA for induction of remission. In the METEOR trial, all patients were on 10–40mg/day of corticosteroids with or without active disease. The primary outcome was corticosteroid-free remission between weeks 12–30. On meta-analysis, there was no significant difference in rates of inducing remission with methotrexate compared to placebo (RR, 1.31; 95% CI, 0.89–1.94) (eFigure 15). The certainty of evidence was rated as very low due to very serious indirectness (different dosing regimens and modes of administration, variable definition of clinical remission, and inability to truly assess whether remission was induced by corticosteroids or methotrexate), and serious imprecision (Table 12). For maintenance of remission, two trials compared methotrexate to placebo and one compared methotrexate to 5-ASA. Similar to induction, there was no difference between methotrexate and placebo or 5-ASA for maintenance of remission (RR, 1.01; 95% CI, 0.79–1.29) (eFigure 15). The certainty of evidence was rated as very low due to serious indirectness, and very serious imprecision (Table 12).

Short- and long-term side effects of thiopurines and methotrexate are well-recognized including risk of bone marrow suppression, hepatotoxicity (including liver fibrosis with long-term use of methotrexate), non-melanoma skin cancer and lymphoma, especially with thiopurines⁷⁸. Besides the direct risks associated with these therapies, risks associated with use of ineffective therapies and delay in initiation of more effective therapies also need to be considered when evaluating potential harms of intervention.

Thiopurines have a slow onset of action, so they have conventionally been used as maintenance agents rather than as induction agents. In the trials of thiopurine therapy in patients with active UC, outcomes were usually assessed 26 weeks or beyond (in 4 trials), in contrast to more recent trials and clinical practice of induction therapy where response to induction therapy is generally assessed within 8–12 weeks. Real-world cohort studies have confirmed the effectiveness of thiopurines in maintaining steroid-free remission and reducing the risk of colectomy in patients with UC⁷⁹. Thiopurines are inexpensive and convenient to take as oral medications and may be particularly useful for maintaining long-term remission in resource-limited settings. Methotrexate, particular subcutaneous dosing, may be effective for inducing and maintaining remission in patients with CD; the reason for its lack of efficacy in patients with UC is not well understood. Of note, in the contemporary era of advanced therapies, recruitment to trials of methotrexate was challenging with METEOR recruiting 111 patients over 6 years (2007–13) and MERIT-UC recruiting 84 patients over 5 years (2012–16).

We did not identify any new RCT that provided direct evidence for combination immunomodulator-TNF antagonist therapy in moderate-to-severe UC since the last guideline⁵⁶. The UC-SUCCESS trial provides direct evidence for this recommendation⁸⁰. This RCT randomized patients with moderate-to-severe UC to receive infliximab alone (n=77), azathioprine alone (n=76), or a combination of both agents (n=78). At week 16, the primary endpoint of corticosteroid-free clinical remission was achieved by 39.7% of patients receiving combination therapy compared to 22.1% of infliximab treated patients and 23.7% of azathioprine treated patients. A higher rate of mucosal healing was observed in the combination therapy group (62.8%) compared to infliximab (54.6%) or azathioprine (36.8%). This translated to an absolute difference in clinical remission rate of 170 per 1000 for combination therapy compared to infliximab alone (Table 13) and 159 per 1000 for combination therapy compared to immunomodulator alone (Table 14).

In the UC-SUCCESS trial, there was no difference in the rate of serious infections between the combination therapy group compared to those receiving infliximab monotherapy (OR, 0.33; 95% CI, 0.01–7.86). In observational studies, there was a higher risk of lymphoma with combination therapy when compared to TNF antagonist monotherapy. In a national study from France, the risk of lymphoma was higher in patients receiving combination therapy compared to TNF-antagonist monotherapy (HR, 2.53; 95% CI, 1.35–4.77) or thiopurine monotherapy (HR, 2.35; 95% CI, 1.31–4.22)²⁵. Thus, the increase in efficacy with combination therapy must be balanced against the risk of therapy-related adverse events. Patients, particularly with more moderate (rather than severe) disease, may reasonably elect to use TNF antagonist monotherapy over combination therapy balancing risks and benefits.

In addition to direct evidence from the UC-SUCCESS study, indirect evidence in support of combination therapy is also derived from moderate-to-severe CD. In the landmark SONIC trial of immunomodulator-naïve patients with CD, a combination of azathioprine and infliximab therapy was associated with higher rates of corticosteroid-free clinical remission at week 26 compared to those receiving infliximab or azathioprine alone⁸⁰. While the COMMIT trial did not demonstrate benefit of methotrexate added to infliximab in patients with CD, patients receiving combination therapy had a higher infliximab trough level and lower rates of anti-infliximab antibody positivity⁸¹. Thus, it is reasonable to infer, given their similar broad mechanisms of immunosuppression, that there is an additive benefit to use of methotrexate in combination with TNF antagonists. There are less direct data demonstrating the benefit of combination therapy with adalimumab or golimumab. In the DIAMOND trial, a combination of azathioprine and adalimumab treatment was associated with higher rates of endoscopic response compared with adalimumab monotherapy⁸². Given similar rates of immunogenicity across the different TNF antagonists, we extrapolated the benefit of combination therapy to all agents within this class. It has been hypothesized that achieving adequate biologic trough levels may reduce the need for combination therapy in patients receiving TNF antagonist therapy⁸³. This has not been examined prospectively. Similarly, there may be greater benefit to combination therapy in patients genetically predisposed to developing anti-drug antibodies to TNF antagonists⁸⁴.

Since the last guideline in 2020, multiple observational studies and indirect evidence from one randomized trial provided evidence for this recommendation. In a multi-center study of patients with IBD initiating vedolizumab or ustekinumab, there was no difference in remission rates at one year between those on combination therapy vs. monotherapy⁸⁵. A meta-analysis that included 33 studies (6 RCT and 28 cohort studies) identified no difference in clinical outcomes in patients with combination therapy with vedolizumab (OR, 0.84; 95% CI, 0.68–1.05) or ustekinumab (OR, 1.1; 95% CI, 0.87–1.38)⁸⁶. In contrast to this data, in the VIEWS trial that randomized patients with UC on combination vedolizumab and thiopurine treatment to continuation of dual therapy or withdrawal, cessation of thiopurine was associated with no difference in clinical relapses (p=0.12) but a modestly higher fecal calprotectin (p=0.003) and lower rates of histologic remission (p=0.03)⁸⁷.

There are limited data on the safety of combination immunomodulator and non-TNF antagonist biologic therapy. Subgroup analyses from RCTs as well as data from observational studies do not suggest an increase in risk of serious infections or malignancy with combination therapy for non-TNF antagonist biologic therapy. However, these studies have mostly lacked longer-term follow-up data.

Mechanistically, there are several reasons why combination therapy may be less necessary with non-TNF antagonist biologic therapies. First, both vedolizumab and anti-interleukin agents have lower immunogenicity than TNF antagonists^{88, 89}. In addition, for vedolizumab, receptor saturation is already noted at low trough saturations. However, immunomodulators may have an independent benefit especially for preventing relapse. Adding immunomodulators to non-TNF antagonist biologics, particularly in patients who are naïve to immunomodulators, may possibly be beneficial. Besides efficacy, the guideline panel also reviewed that there is an increase in risk of infections and malignancy (lymphoma and non-melanoma skin cancers) with thiopurine use. In view of the limited and conflicting data, the guideline panel identified this area as a knowledge gap to be informed by further studies.

Several RCTs and observational studies have examined the impact of withdrawal of thiopurine therapy in patients on combination immunomodulator and biologic therapy. A systematic review by Katibian et al. summarized the impact of withdrawal of immunomodulators in patients with IBD receiving combination therapy⁹⁰. A total of 8 RCTs comprising 733 patients were identified; three-quarters of the patients in the studies had CD. Most studies required patients to be in sustained corticosteroid-free remission for at least 6 months prior to attempting drug withdrawal. Compared to patients continuing combination therapy, there was no increase in risk of relapse among patients stopping combination therapy (16.8% vs. 14.8%; RR, 1.15; 95% CI, 0.75–1.76) (eFigure 16). The overall certainty of evidence was rated as very low, due to serious risk of bias, indirectness since most patients in these trials had Crohn’s disease and very serious imprecision (Table 15). In contrast to immunomodulator cessation, cessation of TNF antagonists in patients on combination therapy is associated with a two-fold increase in risk of relapse compared with continuing combination therapy (31.5% vs. 11.2%; RR, 2.35; 95% CI, 1.38–4.01) (eFigure 17). The overall certainty of evidence was rated as very low, due to serious risk of bias, indirectness since most patients in these trials had Crohn’s disease and serious imprecision due to low number of events (Table 15). In the STORI trial in CD, nearly 80% of patients ceasing TNF antagonist therapy experienced a disease relapse requiring re-initiation of the biologic within 7 years after treatment cessation⁹¹.

As outlined in the section above, combination therapy with an immunomodulator and TNF antagonist is associated with an increase in risk of infections and lymphoma when compared to monotherapy with either agent alone. Any benefit of combination therapy should be weighed against a potential increase in risk of treatment-related adverse outcomes.

The guideline panel used data from both CD and UC to inform this recommendation as it was felt that the likely expected outcomes are similar across diseases and conclusions can be extrapolated. Most clinical trials of therapy discontinuation examined outcomes with less than two years of follow-up; the panel recognized that the clinical impact of treatment discontinuation may need to be viewed over a longer time horizon. In three trials, infliximab trough concentrations at the end of trial were lower and proportion of patients with antibodies to infliximab was higher in patients who underwent immunomodulator withdrawal vs. those who continued combination therapy. Whether the lower trough level on withdrawal would result in increased rates of clinical relapse with longer follow-up is unknown. In contrast, in the DIAMOND2 trial, mean adalimumab trough concentration and anti-adalimumab antibody positivity rate was not different between patients who had immunomodulator withdrawal compared with those who continued combination therapy⁸². Recognizing the gaps in data and the need for longer follow-up, the panel made no recommendation in favor of or against withdrawal of immunomodulators in patients on a combination of TNF antagonist and immunomodulators. In patients with moderate (as opposed to severe) disease, on their first biologic, long-duration of remission, or at higher risk for treatment related adverse effects of infection or malignancy (such as older adults), it may be reasonable to consider discontinuation of IMM therapy after 12–18 months while continuing TNF antagonist use. It is important to measure TNF antagonist trough levels prior to immunomodulator withdrawal. Patients with higher trough levels are more likely to maintain clinical remission following immunomodulator withdrawal than patients with borderline or low TNF antagonist levels⁹². It is also important to closely monitor for disease relapse using clinical symptoms, biomarkers, and endoscopic assessment (12–24 months after therapy withdrawal).

Since the last guideline published in 2020, we did not identify any new RCTs comparing a strategy of upfront use of advanced therapies and/or immunomodulator therapy vs. step-up therapy in patients with moderate-to-severe UC. We also did not identify any trials comparing the efficacy of advanced therapies vs. 5-ASA for patients with moderate-to-severe UC. There are three trials that compared thiopurines in this population^93–95. Based on a meta-analysis of these studies, patients treated with thiopurines achieved higher rates of corticosteroid-free clinical remission compared to patients treated with 5-ASAs. In the UC-SUCCESS trial, while infliximab was not more efficacious than immunomodulator monotherapy for achieving clinical remission, it was more effective for achieving endoscopic improvement⁸⁰. By extension, advanced therapies would be more effective than 5-ASA for induction of remission in patients with moderate-to-severe UC. It is important to note that 5-ASAs are not indicated for the treatment of moderate-to-severe UC and have been approved for mild-to-moderate UC. Based on this indirect evidence, it follows that delaying treatment of moderate-to-severe UC with advanced therapies or immunomodulators to treat with 5-ASA drugs may be detrimental, both because 5-ASAs may not work as primary therapy and because use of these drugs will introduce a treatment delay impairing quality of life and increasing risk of complications. Based on serious indirectness of the evidence with unclear estimates of magnitude of benefit, we rated the certainty of evidence as very low.

Risks associated with advanced therapies or immunomodulator therapy have been outlined earlier and may be greater than those associated with 5-ASA therapy. However, these risks should be interpreted in the context of risks of UC-related complications, including colectomy, hospitalization, and persistent disease activity resulting in inferior quality of life, if step-up therapy is used.

Inadequately controlled UC is associated with an increased risk of colectomy, hospitalization, corticosteroid use, and long-term risk of colorectal cancer. UC is a progressive disease that can result in bowel damage⁹⁶. Hence, risk-congruent therapy is warranted to minimize risk of short- and long-term complications. Unfortunately, prediction models to identify patients at high risk of complications or ‘disease severity’ indices have not been well-validated. Ideally, evidence regarding top-down vs. step-up therapy would be best informed by a pragmatic RCT comparing outcomes in patients assigned to risk-congruent therapy vs. conventional management. In the absence of these data, based on indirect evidence, it is likely that step-up therapy using 5-ASAs first, particularly in patients on the more severe side of the disease spectrum with more severe disease, may be detrimental.

Since the last guideline published in 2020, we did not identify any new RCTs directly addressing withdrawal of 5-ASA therapy in patients with moderate-to-severe UC being treated with immunomodulators or advanced therapies. Mantzaris et al. randomized patients with moderate-to-severe UC, in corticosteroid-free clinical, endoscopic and histologic remission on azathioprine and olsalazine, to either continuing azathioprine and olsalazine (0.5mg thrice daily) or azathioprine alone⁹⁷. Over the course of two years, there were no observed differences in risk of relapse severe enough to merit corticosteroid use (RR, 1.02; 95% CI, 0.77–1.34).

To examine whether concomitant 5-ASA impacts treatment response with advanced therapies, we updated our prior approach, conducting a pooled analysis of individual patient-level data of RCTs of advanced therapies in patients with moderate-to-severe UC. By design, all patients in these trials had moderate to severely active disease, despite prior or concomitant 5-ASA exposure. The patients in these trials had to maintain their baseline medications, so they could not stop or start 5-ASA during the trial. Across ten RCTs with 6,044 patients, 4,134 patients received concomitant 5-ASA at baseline and maintained stable dose throughout the induction period. Compared with patients not receiving 5-ASA, patients receiving concomitant 5-ASA were slightly older, more likely to be non-White, more likely to have moderate endoscopic activity, have shorter disease duration, slightly lower C-reactive protein, and more likely to be biologic-naïve. Subsequently, we compared the relative risk of active intervention vs. placebo in patients who were vs. not on concomitant 5-ASA during the trial, using extended modified Poisson regression model with studies being considered as clusters. The model contained main effects of drug, exposure or non-exposure to concomitant 5-ASA, their product term, and confounders including prior biologic exposure, endoscopic severity at baseline, race, disease extent, baseline CRP, fecal calprotectin, hemoglobin, albumin, concomitant corticosteroids and smoking status. After adjusting for confounding variables, the ratio of RR for inducing clinical remission in those on concomitant 5-ASA vs. no concomitant 5-ASA was 1.04 (95% CI, 0.78–1.39), suggesting no treatment effect modification. The overall certainty of evidence supporting lack of benefit of continuing vs. stopping 5-ASAs in patients with moderate-to-severe UC being treated with advanced therapies, after prior exposure to and failure of 5-ASA, was rated as low. Evidence was rated down due to imprecision and indirectness (Table 16).

5-ASAs are generally safe medications, with very low rates of idiosyncratic serious or life-threatening complications⁹⁸. There are rare reports of allergic interstitial nephritis, pancreatitis, pericarditis, myocarditis, and pneumonitis. Continuing 5-ASA may add pill burden and contribute to high cost of care.

We relied on a combination of direct evidence in patients on thiopurines, and indirect evidence in patients treated with advanced therapies, to determine the efficacy of continuing vs. stopping 5-ASA in patients who escalate therapy after failing 5-ASA. Due to the short duration of follow-up in clinical trials, we were not able to study the impact of concomitant 5-ASAs on longer-term risk of disease-related complications including surgery and development of colorectal neoplasia. Large observational studies have suggested no difference in the risk of adverse clinical outcomes between patients who continue vs. stop 5-ASA after starting advanced therapies or immunomodulators^99–101. The guideline panel acknowledged that there may be a subset of patients with UC who have improved, but not achieved remission, with advanced therapies and who continue to experience mild to moderate symptoms due to residual proctitis – these patients may benefit from continuing or adding rectal 5-ASA. The guideline panel also acknowledged that one proposed benefit of long-term 5-ASA use is potential chemoprevention effect against colorectal cancer, but this remains unproven. While large observational studies and meta-analyses have variably suggested that UC patients treated with 5-ASA have lower risk of developing colorectal cancer, recent evidence suggests that chronically active disease is a strong risk factor for developing neoplasia, and sustained remission is a protective factor against colorectal cancer regardless of therapy used that achieves this outcome¹⁰².

The guideline panel identified several knowledge gaps that need to be addressed in future studies.

Recognizing the rapid evolution of drug development and transforming treatment strategies, the AGA will update relevant recommendations from these guidelines by periodic review of evidence every six months. The evidence reviewed will include availability of Phase 3 or Phase 4 efficacy data for new treatments, treatment strategies or existing treatments, as well as significant new safety concerns informing treatment positioning.