Comparative Risk of Serious Infections with Biologic Agents and Oral Small Molecules in Inflammatory Bowel Diseases: 2026 Update

The expanding armamentarium of advanced therapies for inflammatory bowel disease (IBD) has placed a premium on optimal positioning and sequencing. Comparative efficacy is increasingly well characterized, but comparative safety is not: randomized controlled trials (RCTs) are underpowered for infrequent events, are of short duration, and enroll selected populations. Our 2023 systematic review of 20 head-to-head studies predated comparative data for selective interleukin (IL)-23 antagonists and Janus kinase (JAK) inhibitors.¹ We updated that review to compare the risk of serious infections across all approved advanced-therapy classes, stratified by disease phenotype.

We searched Ovid MEDLINE, EMBASE, and Web of Science from February 28, 2022 (the end date of the prior review) through May 1, 2026, supplemented by bibliography, trial registry, and conference searches. We included RCTs and cohort studies of adults with IBD (population) that compared an approved advanced therapy — TNF antagonists, vedolizumab, IL-12/23 antagonists (ustekinumab is the only approved agent in this class, and the two terms are used interchangeably hereafter), selective IL-23 antagonists, or JAK inhibitors (intervention) — with an alternative advanced therapy (comparator), and reported the risk of serious infections (outcome), with at least 100 patients per treatment arm. Serious infection was defined as an infection requiring hospitalization, intravenous antimicrobial therapy, or discontinuation of immunosuppressive therapy, or resulting in death; for RCTs, serious infections coded per the Medical Dictionary for Regulatory Activities were accepted. Most fully adjusted estimates were pooled using DerSimonian–Laird random-effects models, with the hazard ratio (HR) as the primary effect measure. Subgroup analyses by phenotype (Crohn’s disease [CD] vs ulcerative colitis [UC]) and age, and meta-regression on six study-level covariates, were prespecified. Full methods are provided in the Supplementary Methods.

We screened 5,143 records, reviewed 178 full-text articles, and identified 25 newly eligible studies; 15 of the 20 studies in the prior review were carried forward. The final dataset comprised 40 studies — 5 RCTs and 35 observational cohorts — enrolling 326,403 patients (Supplementary Table 1). Fifteen studies were conducted in North America and 11 in Europe, and 11 were multinational; most cohorts used a new-user design with propensity-score or multivariable adjustment. Median age was 42 years (interquartile range [IQR], 40–44), 50% of patients were female (IQR, 46–53), and 37% were receiving corticosteroids at baseline (IQR, 27–52); follow-up ranged from 6 to 49 months. Across comparisons, median incidence rates of serious infections ranged from 3.06 to 4.40 per 100 person-years. Cohort studies were at low-to-moderate risk of bias (Newcastle–Ottawa score 7–9) and all RCTs were at low risk.

In patients with CD (Figure 1A), ustekinumab was associated with a lower risk of serious infections compared with TNF antagonists (HR, 0.84 [95% CI, 0.79–0.89]; 7 studies; I²=46%) (Supplementary Figure 1A) and compared with vedolizumab (0.67 [0.52–0.86]; 10 studies; I²=61%) (Supplementary Figure 1B). Selective IL-23 antagonists did not differ from TNF antagonists (1.00 [0.68–1.47]; 1 study),² vedolizumab (1.05 [0.71–1.54]; 1 study),² or ustekinumab (1.04 [0.75–1.43]; 4 studies; I²=0%); in the three RCTs informing the last comparison, serious infections occurred in 8/255 vs 11/265 (SEQUENCE),³ 14/630 vs 9/309 (VIVID),⁴ and 3/595 vs 0/300 (GALAXI).⁵ Vedolizumab was not associated with a lower risk than TNF antagonists (1.06 [0.96–1.16]; 11 studies; I²=48%) (Supplementary Figure 1C). In contrast, JAK inhibitors were associated with a higher risk than both TNF antagonists (1.36 [1.05–1.77]; 3 studies; I²=0%) (Supplementary Figure 1D) and vedolizumab (1.26 [1.05–1.51]; 2 studies; I²=0%), but did not differ from selective IL-23 antagonists (0.98 [0.49–1.95]; 1 study).²

Figure 1. Comparative risk of serious infections with advanced therapies in inflammatory bowel disease, stratified by disease phenotype.

Forest plots of pooled adjusted hazard ratios (HR) for serious infections across pairwise comparisons of advanced therapies in (A) Crohn’s disease and (B) ulcerative colitis, derived from random-effects meta-analyses of the phenotype-stratified data. The number of contributing studies is shown for each comparison; single-study rows in panel A are from Park et al 2026 (Crohn’s disease multi-drug cohort). Filled squares denote statistically significant estimates and open squares nonsignificant estimates; horizontal lines denote 95% confidence intervals. HR <1 favors the first-named (intervention) agent. IL, interleukin; JAK, Janus kinase; TNF, tumor necrosis factor. No separable ulcerative colitis estimate was available for selective IL-23 antagonists. Study-level forest plots with individual study weights and I² statistics for four key pairwise comparisons are shown in Supplementary Figure 1.

In patients with UC (Figure 1B), both vedolizumab (0.88 [0.81–0.96]; 12 studies; I²=47%) (Supplementary Figure 1C) and ustekinumab (0.74 [0.66–0.84]; 3 studies; I²=0%) (Supplementary Figure 1A) were associated with a lower risk than TNF antagonists, and ustekinumab with a lower risk than vedolizumab (0.81 [0.71–0.91]; 4 studies; I²=0%) (Supplementary Figure 1B). JAK inhibitors did not differ from TNF antagonists (0.79 [0.59–1.06]; 5 studies; I²=76%) (Supplementary Figure 1D) or vedolizumab (1.08 [0.81–1.44]; 4 studies; I²=66%). No separable UC estimate was available for selective IL-23 antagonists.

Phenotype modified two comparisons: vedolizumab vs TNF antagonists and JAK inhibitors vs TNF antagonists (both P<.01 for interaction), with the substantial heterogeneity seen in the unstratified analyses (I²=53% and 79%, respectively) largely resolving on stratification. Age did not modify any comparative association. Study-level meta-regression, which tests modification of comparative estimates rather than individual risk, identified no significant modifier other than baseline corticosteroid use for JAK inhibitors vs ustekinumab, a comparison in which JAK inhibitors were associated with a higher risk in mixed IBD cohorts (1.41 [1.12–1.76]; 4 studies; I²=0%).^{6, 7}

In this updated meta-analysis of 40 head-to-head studies, the comparative risk of serious infections with advanced therapies in IBD was phenotype-specific. In UC, vedolizumab and ustekinumab were associated with a lower risk than TNF antagonists and JAK inhibitors carried no detectable excess risk; in CD, ustekinumab and selective IL-23 antagonists had the most favorable profile, whereas JAK inhibitors were associated with a higher risk than TNF antagonists and vedolizumab.

These phenotype-specific patterns broadly parallel each agent’s comparative efficacy in the corresponding phenotype.^{9, 10} Whether more effective disease control lowers net infection risk, through reduced corticosteroid exposure and fewer disease-related septic complications, cannot be determined from these data, which lack patient-level information on disease activity and treatment response; confounding by indication, treatment sequence, and calendar time is an equally plausible explanation. This hypothesis requires testing in longitudinal studies linking treatment response to subsequent infection, as in the PYRAMID registry analysis, in which adalimumab responders had a 34% lower risk of serious infection.⁸ Further limitations include reliance on observational data with heterogeneous covariate adjustment, leaving residual confounding by inconsistently captured severity markers (penetrating or fistulizing behavior, prior surgery, cumulative corticosteroid exposure, therapy sequence) as an alternative explanation for the phenotype-specific findings; non-standardized outcome definitions; and sparse comparative data for selective IL-23 antagonists and for several JAK inhibitor pairings, which rest on one or two cohorts. These estimates are therefore best interpreted as individual pairwise comparisons rather than as a ranking of therapies and weighed alongside comparative efficacy when selecting therapy for an individual patient.

Supplementary Material

The 40 studies included in the systematic review are listed in the Supplementary References.

Data availability:

All data supporting the findings of this study are derived from published studies listed in the Supplementary References and are presented in full in the manuscript and Supplementary Material. The data abstraction file and analytic code are available from the corresponding author on reasonable request.