Meta-analysis: safety and efficacy of ustekinumab in pediatric inflammatory bowel disease patients with anti-TNF failure

Abstract

Background & aims

Evidence for ustekinumab (UST) in pediatric inflammatory bowel disease (IBD) is limited, particularly in patients who do not respond to anti–tumor necrosis factor (anti-TNF) therapy. We conducted a meta-analysis to evaluate the efficacy and safety of UST in pediatric IBD after anti-TNF failure.

Methods

PubMed, Cochrane Library, Embase, and Web of Science were searched from inception through August 2025. We included studies reporting outcomes of UST in pediatric IBD patients with prior anti-TNF failure. Pooled estimates were calculated using fixed- or random-effects models according to heterogeneity (I²).

Results

Fifteen studies including 653 pediatric patients were analyzed. Pooled clinical remission rates were 57% (95% CI, 28%–86%) at 8–16 weeks and 68% (95% CI, 58%–79%) at week 52. Steroid-free clinical remission was 44% (95% CI, 38%–49%) at 8–16 weeks and 58% (95% CI, 45%–70%) at week 52. The pooled endoscopic remission rate was 42% (95% CI, 35%–49%) and a biochemical remission rate of 63% (95% CI, 40%–87%). Adverse events occurred in 20% of patients (95% CI, 9%–30%), and serious adverse events occurred in 1% (95% CI, 0%–2%). At week 52, clinical remission rates were 69% in Crohn’s disease (CD) and 65% in ulcerative colitis (UC).

Conclusion

In pediatric IBD patients with prior anti-TNF failure, UST was associated with favorable pooled remission rates and an acceptable safety profile, although substantial inter-study heterogeneity warrants cautious interpretation. These findings support UST as a feasible therapeutic option in this refractory pediatric setting; while highly generalizable to pediatric CD, the findings for the UC subgroup offer important preliminary evidence that warrants cautious interpretation due to limited precision.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12876-026-05078-9.

Introduction

IBD is a group of chronic, recurrent, idiopathic inflammatory conditions that affect the gastrointestinal tract. It primarily comprises CD and UC [1]. On a global scale, there has been a progressive rise in the frequency of IBD diagnoses over the past two decades. In Western countries, statistics indicate that the prevalence can be as high as one in every 200 individuals [2]. Notably, nearly one-quarter of all IBD cases are diagnosed during childhood or adolescence [3]. Over the past three decades, the global burden of pediatric IBD has increased substantially, with both incidence and prevalence showing persistent upward trends and the most pronounced rise observed among younger children. Incidence rates are particularly high in Northern Europe, reaching up to 23.1 per 100,000 person-years, while lower but progressively increasing rates have also been reported in Asia and the Middle East (0.5–21.6 per 100,000 person-years) [4]. Compared to adult-onset forms, pediatric IBD patients face more severe challenges, including more extensive disease, impaired growth and development, delayed puberty and psychosocial difficulties [5].

Current therapeutic strategies for pediatric IBD include dietary and nutritional therapy, 5-aminosalicylic acid (5-ASA) agents, corticosteroids, immunomodulators such as thiopurines or methotrexate, and biologic therapies, particularly anti-TNF agents. However, long-term disease control remains challenging in some patients because of corticosteroid dependence, primary non-response (PNR), and secondary loss of response (LOR) to anti-TNF therapy [6]. Consequently, there is a pressing need to identify alternative therapeutic options capable of overcoming treatment resistance in pediatric patients. In this context, biologic agents targeting the interleukin-12/interleukin-23 (IL-12/IL-23) pathway have emerged as promising treatments for patients with refractory disease [7].

UST, a fully human monoclonal antibody targeting the shared p40 subunit of IL-12 and IL-23, was the first agent approved for clinical use within this therapeutic class. The results of the UNITI-1 and UNITI-2 trials for Crohn’s disease and the UNIFI trial for ulcerative colitis have demonstrated the efficacy and safety of UST in both induction and maintenance therapy for adult IBD [8, 9]. While substantial evidence supports the effectiveness of UST in adult populations, data regarding its use in pediatric IBD remain limited, highlighting an unmet need for robust evaluation of its safety and efficacy in this younger population [10].

Furthermore, due to the lack of high-quality prospective studies, clinical evaluations of UST in this refractory pediatric population remain limited, with most of the existing evidence derived from retrospective cohort studies. Therefore, systematically synthesizing and quantitatively analyzing these scattered real-world data will help provide a more comprehensive summary of the current evidence and offer an evidence-based foundation for clinical decision-making.

Methods

This research constitutes a meta-analysis registered on the PROSPERO platform [11] (ID: CRD420251045145) and adheres to the PRISMA guidelines [12]. This systematic review and meta-analysis were conducted and reported in accordance with the 2020 version of the PRISMA guidelines. More details are provided in Table S2. This study follows the PICO framework: the study population (P) comprises pediatric patients aged < 18 years with IBD who experienced failure of prior anti-TNF therapy; the intervention (I) refers to ustekinumab (UST) treatment; the comparison (C) is absent, as all included studies were single-arm observational designs; the outcomes (O) cover efficacy endpoints (clinical remission rate, corticosteroid-free remission rate, endoscopic remission rate and biochemical remission rate) and safety endpoints (adverse events and serious adverse events).

Inclusion and exclusion criteria

To be incorporated into this meta-analysis, studies were required to satisfy the following eligibility requirements: (1) the cohort comprised pediatric individuals (aged under 18) with a diagnosis of IBD based on the updated Porto criteria; (2) At least 80% of the participants had received at least one TNF antagonist treatment before starting ustekinumab therapy; (3) The studies provide clear data on efficacy and safety. Additionally, we excluded the following types of literature: (1) randomized controlled trials (RCTs); (2) conference abstracts, editorials, or other non-peer-reviewed publications; (3) a total sample size of fewer than 10 patients; (4) insufficient or incomplete data that precluded extraction or analysis.

The aim of this study is to evaluate the efficacy and safety of UST in pediatric patients who have failed anti-TNF therapy in real-world clinical settings; therefore, we focused on including real-world observational cohort studies. Although RCTs have high internal validity, their strict inclusion and exclusion criteria may not fully reflect the dose adjustments, shortened dosing intervals, and complex treatment contexts commonly observed in pediatric refractory IBD. In addition, combining data from strictly controlled RCTs with data from single-arm observational cohorts may introduce significant methodological heterogeneity.

Search strategy

We conducted a comprehensive electronic search across multiple databases, including PubMed, Embase, Web of Science, and Cochrane Library, from the establishment of the databases until August 13, 2025. The search methodology integrated both Medical Subject Headings (MeSH) and relevant keywords, tailored to the specific syntax and indexing requirements of each platform. Our main search terms included terms related to “ustekinumab” and “inflammatory bowel disease” (including “Crohn’s disease” and “ulcerative colitis”), as well as keywords related to the “pediatric” population. The detailed search strategy for each database is provided in Table S1.

Study selection and data extraction

The identification of eligible literature was carried out independently by two researchers, who applied the PICO framework based on our established inclusion and exclusion benchmarks. Any disagreements that arose during the selection phase were adjudicated through mutual deliberation or, if needed, by seeking input from a third reviewer. Data extraction was performed using EndNote software, and relevant information from the included studies was systematically organized. Extracted variables included first author, year of publication, geographic region and country, patient age, proportion of patients with prior anti–TNF therapy exposure, study design, IBD subtype, sample size, and treatment regimen. We also extracted data on treatment durability, including UST continuation, discontinuation, loss of response, or treatment failure; concomitant or bridging therapies, and the reported timing of clinical response or remission. All extracted data were subsequently entered into a standardized Microsoft Excel spreadsheet for further analysis.

Study quality assessment

Two investigators independently assessed the risk of bias of single-arm studies using the Methodological Index for Non-Randomized Studies (MINORS) scale [13]. The MINORS instrument comprises 12 methodological items, of which the first eight are applicable to non-comparative studies and the remaining four to comparative studies. As only single-arm studies were included in this meta-analysis, assessment was restricted to the first eight items. Each item was scored from 0 to 2, yielding a maximum possible score of 16 points. A score of 0 indicated that the item was not reported, 1 indicated inadequate reporting, and 2 indicated adequate reporting. Any discrepancies between the two reviewers were resolved through discussion or, if necessary, by consultation with a third reviewer. More information can be found in Table S3.

Statistical analysis

All data synthesis was executed through SPSS AU Web Edition. We calculated pooled event rates for binary outcomes and presented them alongside their 95% confidence intervals (CIs). The I² index was employed to evaluate inter-study heterogeneity, with values of 25%, 50%, and 75% designating low, moderate, and high levels of variability, respectively. Adhering to the Cochrane Handbook’s recommendations, we utilized a fixed-effects model when I²≤50% and P ≥ 0.10, while a DerSimonian-Laird random-effects approach [14] was adopted to address significant heterogeneity (I² > 50% or P < 0.10). The use of a random-effects model was considered appropriate because clinical and methodological heterogeneity was expected across real-world pediatric IBD studies, including differences in disease subtype, baseline severity, prior biologic exposure, dosing regimens, outcome definitions, and follow-up duration. Under this model, the pooled estimate was interpreted as an average effect across heterogeneous study settings rather than as a single common treatment effect. Therefore, for outcomes with substantial heterogeneity, pooled estimates were regarded as descriptive summaries of real-world evidence and were interpreted with caution. To ensure the reliability of our findings, sensitivity analyses were performed by the leave-one-out method for outcomes with substantial heterogeneity. Finally, potential publication bias was scrutinized via Egger’s regression analysis and the visual inspection of funnel plots [15].

Results

Our systematic search across electronic databases initially yielded 916 citations. After eliminating 419 duplicate articles, we evaluated the remaining 497 articles. After preliminary screening of the titles and abstracts, we excluded 418 articles, and finally retained 79 articles for full-text evaluation. Among them, 64 articles were excluded because they did not meet the inclusion criteria. In total, 15 studies [10, 16–29] were included in this meta-analysis. The stepwise selection procedure is visualized in the PRISMA flowchart (Fig. 1).

Fig. 1.

PRISMA flow chart of the study illustrating the study selection process

Altogether, 15 observational cohort studies were analyzed, involving a pooled population of 653 pediatric IBD patients (560 with CD and 93 with UC) who had previously failed anti-TNF treatment. Because CD patients accounted for the majority of the pooled population, the overall estimates of this meta-analysis should be understood as being driven primarily by pediatric CD data. In the included literature, the median age of the subjects ranged from 9.5 to 16.4 years. Regarding the treatment history, each study reported that more than 80% of the participants had received anti-TNF drug treatment, and 10 of the studies [10, 17, 19, 20, 23–27, 29] stated that all patients had a history of anti-TNF drugs. A comprehensive overview of these study characteristics is presented in Table 1.

Table 1.

Characteristics of included studies

Author	Country	n	Age	Design type	IBD type	Prior ≥1TNF (%)	UST dose	MINORS (points)	Outcome type
Mitchel2025 [20]	USA	101	16.4(1.8–18.1)	CH	CD	86%	Induction: IV, 43–55 kg:260 mg;55–85 kg:390mg;>85 kg:520mg;<40 kg:6 mg/kg Maintenance: SQ, 90 mg q8w(n = 83); q4w(n = 4); q6w(n = 4)	13	②④⑥⑦⑧
Alhadab2024 [16]	Saudi Arabia	13	13(11.5–14)	CH	CD, UC	100%	Induction: IV, < 40 kg: 9 mg/kg; >40: 390 mg Maintenance: SQ, < 40 kg: 2–4 mg/kg; >40 kg: 90 mg	12	②④⑤⑦⑧
Yamamoto2024 [17]	Japan	43	13.6(8–16)	CH	CD	100%	Induction: IV, < 55 kg: 260 mg; 55–85 kg: 390 mg; >85 kg: 520 mg. Maintenance: SQ, 90 mg, q8-12w	13	①②③④⑤⑦⑧
Pujol-Muncunill2024 [18]	Multi-country	101	15.4(12.7–17.2)	CH	CD	99%	Induction: IV, < 55 kg: 260 mg; 55–85 kg: 390 mg; >85 kg: 520 mg. Maintenance: 90 mg, q8w (79%); 90 mg, q12w (13%)	13	③④⑦⑧
CohenA2021[19]	Canada	11	15.3 ± 1.8	CH	CD	100%	Induction: SQ, 90 mg (w0,1,2) (n = 6); 270 mg (w0), 180 (w1,2) (n = 1); IV,4.5-7.5 mg/kg Maintenance: SQ, 90 mg, q8w	10	⑤⑦⑧
Kim2021[20]	USA	38	12.5(10.2–15)	CH	CD	100%	Induction: IV, 260 mg (n = 18); 390 mg (n = 16); 520 mg (n = 4) Maintenance: SQ, 90 mg. q8w	13	②⑦⑧
Takeuchi2020 [21]	Japan	17	13.6(8–16)	CH	CD	94.1%	Induction: <55 kg: 260 mg; 55–85 kg: 390 mg; >85 kg: 520 mg. Maintenance: 90 mg, q8-12w	12	④⑤⑦⑧
Dayan2019 [22]	USA	52	16.8(14–18)	CH	CD	81%	Induction: <55 kg: 260 mg (n = 20); 55–85 kg: 390 mg (n = 25); >85 kg: 520 mg (n = 2); SQ (n = 5) Maintenance: 90 mg, q8w	13	②④⑦⑧
Koudsi2023 [23]	France	53	15.03(8.7–18)	CH	CD, UC	100%	Induction: IV, 6 mg/kg Maintenance: 90 mg, q8w	13	⑦⑧
CohenS2024 [24]	Multi-country	58	14.5(11.5–16.5)	CH	UC	100%	NR	13	①②③④⑤⑥⑦⑧
Do2021 [25]	USA	10	9.5(2–14)	CH	CD	100%	Induction: 90 mg × 3 q4w (n = 1); 45 mg × 2 (n = 1); 180 mg (n = 1); 240 mg (n = 1); 260 mg (n = 2); 390 mg (n = 3); 520 mg (n = 1) Maintenance: 45 mg, q8w (n = 1); 90 mg, q4w (n = 1); 90 mg, q6w (n = 1); 90 mg, q8w (n = 6); 45 mg × 1, then 90 mg, q6w (n = 1)	11	②⑦⑧
Rehman2024 [26]	USA	10	14.3(11.5–17.5)	CH	UC	100%	Induction: IV, < 55 kg: 260 mg; 55–85 kg: 390 mg; >85 kg: 520 mg Maintenance: SQ, 90 mg, q8w	13	④⑤⑥⑧
Chavannes2018 [27]	Multi-country	44	16(13–17)	CH	CD	100%	Induction: SQ, 90 mg/week × 3 weeks. Maintenance: SQ, 90 mg, q8w (n = 31); q4w (n = 13)	13	①②④⑦⑧
Yerushalmy-Feler2022 [28]	Multi-country	69	15.8(13.8–16.9)	CH	CD	98.6%	Induction: IV, 6 mg/kg Maintenance: SQ, 90 mg, q8w (n = 67); q12w (n = 2)	13	③④⑤⑦⑧
Rébus2024 [29]	France	33	10.7(9.2–12.6)	CH	CD, UC	100%	Induction: IV, < 55 kg: 260 mg; 55–85 kg: 390 mg; >85 kg: 520 mg. Maintenance: SQ, < 45 kg: 45 mg; ≥45 kg: 90 mg q8-12w	11	③④⑦⑧

Abbreviations: MINORS Methodological Index for Non-Randomized Studies, CH Cohort study, NR Not reported, IV Intravenous, SQ Subcutaneous, q4w every 4 weeks, q6w every 6 weeks, q8w every 8 weeks, q12w every 12 weeks, n sample size

①: overall clinical remission rate at 8–16 weeks; ②: overall clinical remission rate at week 52; ③: overall steroid-free clinical remission rates at 8–16 weeks; ④: overall steroid-free clinical remission rate at week 52; ⑤: endoscopic remission rates (last follow-up) ; ⑥: biochemical remission rates at weeks 52–56; ⑦: adverse event rates (last follow-up); ⑧: serious adverse event rates (last follow-up)

Treatment course and durability

Available data on follow-up windows, treatment durability, concomitant therapies, dose escalation and timing of response or remission were extracted when reported and are summarized in Table S4. Across the included studies, treatment durability was reported in different ways, including continuation of UST, treatment discontinuation, treatment failure, loss of response, and dose optimization during follow-up. Discontinuation was most commonly attributed to primary non-response, secondary loss of response, persistent active disease, surgery, adverse events, or non-medical factors. Concomitant therapies were variably reported and included corticosteroids, immunomodulators, 5-aminosalicylic acid, enteral or partial enteral nutrition, and, in selected cases, concomitant biologic therapy. Most studies assessed response or remission at predefined follow-up time points, commonly weeks 8–16, week 52, rather than reporting formal time-to-event outcomes. Because these variables differed substantially in definition and reporting across studies, they were summarized narratively and were not included in the pooled analyses.

Efficacy outcome

Primary outcome: clinical remission

Clinical remission was defined as a Pediatric Crohn’s Disease Activity Index (PCDAI) ≤ 10 for CD and a Pediatric UC Activity Index (PUCAI) ≤ 10 for UC. Three studies [17, 24, 27] reported the overall clinical remission rate at weeks 8–16, including both CD and UC patients. Significant heterogeneity was observed among these studies (I² = 93.47%, P < 0.01), necessitating the use of a random-effects model. The pooled analysis indicated that 57% (95% CI, 28%–86%) of pediatric IBD patients with prior anti–TNF therapy failure achieved clinical remission following UST treatment during this period. Given the substantial heterogeneity, this pooled estimate should be interpreted as a descriptive summary of highly variable real-world outcomes rather than as a precise estimate of a uniform treatment effect (Fig. 2a).

Fig. 2.

a Overall clinical remission rate at 8–16 weeks. b Overall clinical remission rate at week 52. c Overall steroid-free clinical remission rate at 8–16 weeks. d Overall steroid-free clinical remission rate at week 52

Eight studies [10, 16, 17, 20, 22, 24, 25, 27] reported the overall clinical remission rate at week 52. Given the substantial heterogeneity (I² = 77.61%, P < 0.01), a random-effects model was applied. The pooled clinical remission rate at week 52 was 68% (95% CI, 58%–79%), but this estimate should be interpreted cautiously because of inter-study variability. (Fig. 2b).

Primary outcome: steroid-free clinical remission

Steroid-free clinical remission was defined as achieving clinical remission without corticosteroid use in the preceding 30 days. Five studies [17, 18, 24, 28, 29] reported the overall steroid-free clinical remission rate for patients treated for 8–16 weeks, including both CD and UC. No significant heterogeneity was observed among these studies (I² = 0%, P = 0.93), and thus a fixed-effect model was applied. The pooled analysis indicated a steroid-free clinical remission rate of 44% (95% CI, 38%–49%) in pediatric IBD patients with prior anti–TNF therapy failure receiving UST during this period (Fig. 2c). Eleven studies [10, 16–18, 21, 22, 24, 26–29] reported the overall steroid-free clinical remission rate at week 52. Given the substantial inter-study heterogeneity (I² = 87.56%, P < 0.01), a random-effects model was applied. The pooled steroid-free clinical remission rate at week 52 was 58% (95% CI, 45%–70%) among pediatric IBD patients with prior anti–TNF failure receiving UST therapy, although the findings should be interpreted cautiously because of the observed inter-study variability. (Fig. 2d).

Secondary outcome: endoscopic remission

Endoscopic remission was established based on a Simple Endoscopic Score for Crohn’s Disease (SES-CD) ≤ 2 for CD patients and a Mayo Endoscopic Score (MES) of 0 for UC patients. Seven studies [16, 17, 19, 21, 24, 26, 28] reported endoscopic remission rates. Given the substantial heterogeneity among studies (I² = 84.38%, P < 0.01), a random-effects model was applied. The pooled endoscopic remission rate was 42% (95% CI, 35%–49%). Given the substantial heterogeneity, this estimate should be interpreted cautiously as a descriptive summary of heterogeneous real-world evidence. (Fig. 3a).

Fig. 3.

a Endoscopic remission rate at the last follow-up. b Biochemical remission rate at weeks 52–56

Secondary outcome: biochemical remission

Biochemical remission was established based on a C-reactive protein (CRP) level below 5 mg/L and a fecal calprotectin (FC) concentration under 250 µg/g. Three studies [10, 24, 26] reported biochemical remission rates at weeks 52–56. Given the substantial inter-study heterogeneity (I² = 87.99%, P < 0.01), a random-effects model was applied. The pooled biochemical remission rate at weeks 52–56 was 63% (95% CI, 40%–87%) Owing to the substantial heterogeneity and the limited number of contributing studies, this finding should be interpreted cautiously. (Fig. 3b).

Safety outcome

Adverse event

The incidence of adverse events was defined as the proportion of patients experiencing any unexpected medical occurrence during treatment or the study period, including infections, headaches, rashes, and similar events. Fourteen studies [10, 16–25, 27–29] reported adverse event rates. Given the substantial heterogeneity among studies (I² = 95.89%, P < 0.01), likely related to differences in adverse event reporting across centers, a random-effects model was applied. The pooled adverse event rate was 20% (95% CI, 9%–30%) (Fig. 4a). These findings should be interpreted cautiously because of the considerable inter-study variability.

Fig. 4.

a Pooled incidence rate of adverse events. b Pooled incidence rate of serious adverse events

Serious adverse event

The incidence of serious adverse events was defined as the proportion of patients experiencing adverse events leading to death, severe infection, intestinal obstruction, hospitalization, prolonged hospital stay, or permanent disability/functional impairment. Fifteen studies [10, 16–29] reported serious adverse event rates, showing low heterogeneity (I² = 34.02%, P = 0.10); therefore, a fixed-effect model was applied. The pooled incidence of serious adverse events in pediatric IBD patients with prior anti–TNF therapy failure treated with UST was 1% (95% CI, 0%–2%) (Fig. 4b).

Subgroup analysis

Subgroup analyses were performed according to the type of IBD (CD vs. UC). For 52-week clinical remission percentage, seven studies [10, 16, 17, 20, 22, 25, 27] reported outcomes in the CD subgroup, and three studies [16, 22, 24] reported outcomes in the UC subgroup. Significant statistical variability was detected in the CD subgroup (I² = 79.85%, P < 0.01), and a random effects model was required for the analysis. The pooled clinical remission rate at week 52 was 69% (95% CI, 56%–82%) for pediatric CD patients with prior anti–TNF therapy failure treated with UST. In the UC subgroup, no heterogeneity was detected (I² = 0%, P = 0.66), allowing use of a fixed-effect model. The pooled clinical remission rate at week 52 was 65% (95% CI, 54%–76%) for pediatric UC patients (Fig. 5a).

Fig. 5.

a Pooled clinical remission rate at week 52 for CD and UC subgroups. b Pooled steroid-free clinical remission rate at week 52 for CD and UC subgroups

Regarding steroid-free clinical remission at week 52, eight studies [10, 17, 18, 21, 22, 27–29] provided data for the CD subgroup, and four studies [22, 24, 26, 29] for the UC subgroup. In the CD subgroup, significant heterogeneity was observed (I² = 89.41%, P < 0.01), and a random-effects model was applied. The pooled steroid-free clinical remission rate at week 52 was 52% (95% CI, 37%–68%) for pediatric CD patients with prior anti–TNF therapy failure treated with UST. In the UC subgroup, heterogeneity was also significant (I² = 78.81%, P < 0.01), warranting a random-effects model. Among the pediatric UC cohort, a synthesized steroid-free clinical remission rate of 60% (95% CI, 36%–83%) was achieved by the 52-week mark (Fig. 5b).

To examine whether pooled outcomes were primarily influenced by studies using intensified ustekinumab regimens, we conducted an exploratory subgroup analysis according to dosing intensity. For corticosteroid-free clinical remission at week 52, the pooled rate was 65% (95% CI, 46%–84%) in the standard/near-standard dosing subgroup and 55% (95% CI, 40%–71%) in the intensified/mixed dosing subgroup. The difference between subgroups was not statistically significant (Q = 0.62, P = 0.43), indicating that dosing intensification did not appear to have a substantial impact on the pooled week-52 corticosteroid-free remission estimate at the study level (Figure S1).

With regard to adverse events, the incidence rate reported in the standard/near-standard dose subgroup was numerically higher than that in the intensive/mixed dose subgroup. The pooled incidence of adverse events was 79% in the standard/near-standard dose subgroup and 9% (95% CI, 5%–12%) in the intensive/mixed dose subgroup, with a significant difference between the two groups (Q = 35.80, P < 0.01). However, this result should be interpreted with caution, as the standard/near-standard dose subgroup included only two studies, both of which reported relatively high rates of adverse events. Furthermore, when using a random-effects model, the confidence interval for the standard/near-standard dose subgroup exceeded the logical upper limit of 100%, further indicating the instability of this estimate. The results should be interpreted as follows: based on currently available study-level data, no signal of a marked increase in adverse events was observed in the intensive/mixed-dose group (Figure S2).

Sensitivity analysis

We conducted sensitivity analyses by sequentially removing each study to assess its impact on the combined estimate, especially for results with an I² value greater than 50%. For the 52-week clinical remission rate, exclusion of the study by Chavannes [27] substantially reduced heterogeneity (I² = 47.48%, P = 0.08), yielding a pooled clinical remission rate of 73% (95% CI, 65%–80%). Regarding endoscopic remission, removing the studies by Alhadab, Rehman, and Yerushalmy-Feler [16, 26, 28] markedly decreased heterogeneity (I² = 0%, P = 0.81), with a pooled remission rate of 25% (95% CI, 12%–37%). For adverse event incidence, exclusion of the studies by Yamamoto, Takeuchi, and Rébus [17, 21, 29] reduced heterogeneity considerably (I² = 24.88%, P = 0.21), resulting in a pooled incidence rate of 9% (95% CI, 7%–12%). After excluding other individual studies, the pooled estimates did not change substantially, suggesting that the overall direction of the findings was broadly stable; however, the remaining heterogeneity should still be considered when interpreting these results.

Publication bias detection

Funnel plots were generated to assess potential publication bias for the primary efficacy and safety outcomes. Figure 6 presents the funnel plots for the following endpoints: 52-week clinical remission rate, 52-week steroid-free clinical remission rate, incidence of adverse events, and incidence of serious adverse events. Visual inspection of these plots did not indicate any substantial publication bias.

Fig. 6.

a: W52 Clinical Remission Funnel Plot b: W52 Corticosteroid-Free Clinical Remission Funnel Plot c: Adverse Event Funnel Plot d: Serious Adverse Event Funnel Plot

Discussion

Treatment for pediatric IBD patients, similar to that for adults, includes aminosalicylates, immunomodulators, anti-TNF agents, and corticosteroids [30]. Despite the expanding array of therapeutic options, anti-TNF agents remain the only biologics approved and demonstrated to be effective for moderate-to-severe pediatric IBD. However, their use carries risks such as allergic reactions, autoimmune complications, primary non-response, or secondary loss of efficacy [31, 32]. In this study, we performed a meta-analysis evaluating UST treatment in pediatric IBD patients who had failed anti-TNF therapy. To our knowledge, this represents the first meta-analysis specifically focusing on UST in this patient population. The results indicate that UST treatment is associated with substantial efficacy and a favorable safety profile in pediatric IBD patients refractory to anti-TNF therapy.

Clinical remission refers to the state in which disease symptoms are well-controlled and the patient maintains overall stability. Achieving clinical remission is particularly important for IBD patients, as it directly influences quality of life, long-term prognosis, and healthcare resource utilization [33]. In the present study, the clinical remission rate among pediatric IBD patients treated with UST for 8–16 weeks was 57%, increasing to 68% by week 52, indicating enhanced long-term efficacy. Previous research has similarly demonstrated that UST maintains sustained efficacy in both CD and UC patients. Long-term safety data, extending up to five years for CD and four years for UC, further support the favorable safety profile of UST [34]. Subgroup analyses revealed that following anti-TNF therapy failure, pediatric CD patients achieved a 69% clinical remission rate at week 52, whereas pediatric UC patients reached 65%. For comparison, in the UNITI-1 trial of anti-TNF-exposed adults, CD patients treated with UST (90 mg every 8 weeks) achieved a 41.1% clinical remission rate at one year [9]. Another trial demonstrated a 67.9% clinical remission rate at week 56 in UC patients who had failed anti-TNF therapy [8, 35], consistent with the findings of the present study. Mechanistically, the therapeutic effect of UST is mediated through the JAK-STAT signaling pathway, the principal downstream pathway of both IL-12 and IL-23. IL-12 primarily activates STAT4, promoting Th1 cell proliferation and function, whereas IL-23 predominantly activates STAT3, supporting Th17 cell responses. This pathway plays a central role in regulating immune cell differentiation, proliferation, and activity. By inhibiting IL-12 and IL-23, UST effectively suppresses intestinal inflammation, thereby enhancing clinical remission rates in IBD patients [36].

The steroid-free clinical remission rate reflects the successful control and sustained absence of disease symptoms without glucocorticoid use, serving as a key indicator for evaluating treatment efficacy and optimizing long-term outcomes in IBD patients [37, 38]. In our meta-analysis, pediatric IBD patients treated with UST following anti-TNF therapy failure achieved a steroid-free clinical remission rate of 44% at weeks 8–16, which increased to 58% by week 52. Subgroup analyses showed steroid-free remission rates of 52% for CD and 60% for UC at week 52. These findings are consistent with a previous study reporting a 51.1% steroid-free remission rate in elderly CD patients (age > 60 years) treated with UST after anti-TNF failure [39], and a meta-analysis indicating a 39% remission rate at one year in steroid-refractory adult UC patients [40]. Rather than reflecting genuine biological differences in drug efficacy between pediatric and adult populations, the apparent superiority of treatment outcomes in children is more likely attributable to methodological differences and biases inherent in the studies. First, the pediatric evidence base consists primarily of retrospective cohort studies with relatively small sample sizes. There is ample evidence that small-scale observational studies are particularly susceptible to selection bias and the small-sample effect, which may lead to an overestimation of treatment efficacy [41]. Second, in real-world pediatric practice, proactive dose-optimization strategies (such as shortening maintenance therapy intervals or increasing doses) are often employed to address the issue of rapid drug clearance in children. Compared to standard maintenance therapy cohorts, such intensive dosing strategies may artificially inflate observed remission rates.

The secondary outcomes included endoscopic and biochemical remission rates. At the final follow-up, pediatric IBD patients with prior anti-TNF failure who were treated with UST achieved an endoscopic remission rate of 42%. Comparable studies in adults reported slightly lower rates: Huang et al. observed a 35.6% endoscopic remission rate in adult CD patients receiving UST after initial biologic therapy between weeks 44 and 56 [42], while Li et al. reported a 39% remission rate at week 44 in adult UC patients treated with UST [43]. These results are broadly consistent with adult findings. However, apparent numerical differences between pediatric and adult cohorts should not be interpreted as evidence of superior efficacy in pediatric patients, as they may reflect differences in study design, patient selection, outcome definitions, follow-up duration, and dose optimization strategies. Mechanistically, UST binds to the p40 subunit shared by IL-12 and IL-23, thereby preventing receptor engagement and blocking downstream signaling. This inhibition reduces proinflammatory cytokines such as IFN-γ and IL-17, helping to control inflammation and promote mucosal repair [9, 44]. Biochemical remission, defined by normalization of CRP and fecal calprotectin levels, was also assessed. In adult populations, Thunberg et al. reported a long-term biochemical remission rate of only 23% in refractory UC [45], and Gonczi et al. reported a rate of 36.6% in refractory CD [46]. In contrast, our data demonstrate a biochemical remission rate of 63% in pediatric IBD patients who failed anti-TNF therapy. This apparent difference may also reflect differences in study design, outcome definitions, follow-up duration, and reporting of biochemical outcomes across studies.

For safety outcomes, we evaluated the incidence of adverse events (AEs) and serious adverse events (SAEs) in pediatric IBD patients. Our analysis revealed that among those who failed anti-TNF therapy, the overall AE rate following UST treatment was 20%, primarily involving infections, perianal lesions, and similar events, while the SAE rate was 1%, including severe outcomes such as death and serious infections. In comparison, a one-year follow-up from the Korean K-STAR cohort reported an AE rate of 28.2% and an SAE rate of 12.7% in adult CD patients treated with UST [47]. Another meta-analysis [48] reported an AE incidence of 13.5% and an SAE rate of 5.6% among IBD patients receiving UST, consistent with our findings. These data collectively support the conclusion that UST is a relatively safe treatment option for pediatric IBD patients who have not responded to anti-TNF therapy. Notably, infections represent the most common adverse event, likely due to UST’s inhibition of the IL-12 and IL-23 pathways, which can transiently suppress immune function and increase susceptibility to opportunistic infections [49].

Variability in real-world dosing paradigms and clinical dose optimization: In pediatric IBD, ustekinumab maintenance interval shortening, such as from q8w to q4w or q6w, and dose intensification are commonly used in real-world practice to address inadequate response, loss of response, or suspected pharmacokinetic underexposure [50]. Several included studies explicitly reported non-standard or intensified regimens, including Chavannes et al. [27], in which a substantial proportion of patients received q4w maintenance, and Yerushalmy-Feler et al. [28], which specifically evaluated a dose-escalation cohort. Therefore, the pooled remission rates in this meta-analysis should not be interpreted as the efficacy of a single fixed standard-dose regimen, but rather as the average outcome of heterogeneous real-world ustekinumab treatment strategies.

To further examine whether the overall findings were primarily driven by intensified dosing, we performed an exploratory subgroup analysis according to dosing intensity. The pooled corticosteroid-free clinical remission rate at week 52 was 65% in the standard/near-standard dosing subgroup and 55% in the intensified/mixed dosing subgroup, with no statistically significant subgroup difference. This finding suggests that the overall efficacy estimate was not simply inflated by studies using intensified dosing strategies. However, because this analysis was based on study-level rather than patient-level dosing data, it cannot establish the independent effect of dose escalation on remission. In addition, dose escalation is likely influenced by disease severity, inadequate response, and clinician decision-making, thereby introducing potential confounding factors related to indications. Furthermore, the subgroup classification was inevitably simplified; the Standard group included only two studies, which may not fully reflect differences in dosing intervals, dose adjustments, or treatment duration across different studies. Therefore, treatment heterogeneity remains an important source of clinical heterogeneity and should be taken into account when interpreting pooled estimates.

It should be noted that our meta-analysis identified significant heterogeneity across multiple measures. Such heterogeneity is to be expected in real-world evidence studies involving pediatric patients with inflammatory bowel disease. The primary drivers of this heterogeneity may include: (1) differences in dosing regimens, including the use of different induction and maintenance treatment regimens (e.g., subcutaneous versus intravenous administration, and varying dose-escalation intervals); (2) variations in endoscopic assessment thresholds and differing definitions of follow-up windows; (3) patient heterogeneity, given the varying baseline disease severity among patients in these cohorts of anti-TNF therapy failures. Therefore, the pooled estimates presented in this paper should be interpreted as a broad descriptive range of real-world outcomes rather than a standardized benchmark of efficacy. In addition, sensitivity analyses demonstrated that exclusion of several influential studies substantially reduced heterogeneity for selected outcomes, indicating that differences in study characteristics and treatment strategies may partially account for the observed variability.

Our study has several limitations. First, most of the studies included were retrospective cohort studies, which meant that important data were not always available at set times and that the reported follow-up intervals varied a lot. Second, the substantial heterogeneity observed in primary and secondary outcomes warrants cautious interpretation of the pooled results. While we utilized random-effects models to account for this variance, the wide confidence intervals observed in several endpoints underscore the need to view these percentages as descriptive estimations of real-world outcomes rather than definitive, universally applicable clinical standards. Third, none of the studies we examined included control groups, which restricted our analysis to single-arm data and may have introduced bias. Another important limitation is that the unequal sample sizes across CD and UC subgroups reflect the current global clinical reality of medication selection. Nevertheless, the independent subgroup analysis for UC remains highly meaningful as it synthesizes scarce real-world evidence. We acknowledge that the UC-specific results have limited precision and should be interpreted with caution. Future disclosure of prospective data specifically targeting pediatric UC is highly anticipated. Fourth, there were discrepancies in the dosing regimens of UST included in the studies. Reporting on patient-level drug exposure, the timing of dose optimization, and dose-related outcome measures varied across studies, making it impossible for us to conduct a formal dose-response analysis. We hope that future studies will further standardize dosing regimens to help elevate the level of evidence. Fifth, it should be noted that a number of pooled estimates have large confidence intervals, especially for short-term clinical remission at 8–16 weeks. This is likely due to significant heterogeneity and small sample numbers across the included studies. Therefore, these findings must be regarded with caution.

In summary, this meta-analysis demonstrates that UST is a feasible and safe treatment option for pediatric patients with IBD who have failed prior anti-TNF therapy. Although our pooled data occasionally show higher efficacy in pediatric patients compared to some adult cohorts, we caution against interpreting this as true biological superiority. Rather, these differences likely stem from heterogeneity in real-world clinical settings—such as small sample sizes, frequent dose optimization, and diverse study designs. Finally, prospective studies with standardized endpoints and dosing strategies are needed to better clarify the efficacy and safety of UST in pediatric IBD following anti-TNF treatment failure.

Abbreviations

UST

Ustekinumab

IBD

Inflammatory bowel disease

CD

Crohn’s disease

UC

Ulcerative colitis

anti-TNF

Anti–tumor necrosis factor

RCT

Randomized controlled trial

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

PICO

Population, intervention, comparison, and outcome

MINORS

Methodological Index for Non-Randomized Studies

IV

Intravenous

SQ

Subcutaneous

q4w

Every 4 weeks

q6w

Every 6 weeks

q8w

Every 8 weeks

q12w

Every 12 weeks

PCDAI

Pediatric Crohn’s Disease Activity Index

PUCAI

Pediatric Ulcerative Colitis Activity Index

SES-CD

Simple Endoscopic Score for Crohn’s Disease

MES

Mayo Endoscopic Score

CRP

C-reactive protein

FC

Fecal calprotectin

AE

Adverse event

SAE

Serious adverse event

CI

Confidence interval

5-ASA

5-aminosalicylic acid

Authors' contributions

Zhaoyang Heng conceived and designed this study, conducted literature retrieval, data extraction and statistical analysis, and wrote the first draft of the thesis. Yuhan Zhang and Keyou Zhou independently screened the relevant research and extracted the data. Li Zhan, Meng Song participated in the data quality assessment and verification. Kelei Zhao, Md Shoykot Jahan and Most Fatema Akter assisted in data interpretation and thesis revision. Ping Lu, Yinghua Ji and Yanting Liu guided the study and reviewed and revised the important academic content of the paper. All authors have reviewed and approved the final draft.

Funding

This study received no funding or financial support.

Data availability

The datasets produced and scrutinized in the current study are available within this article and its corresponding supplementary information.

Declarations

Ethics approval and consent to participate

Not applicable. All patient data used in this study were sourced from previously published studies; therefore, no ethical approval was required.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.