Document resource
Introduction Small bowel capsule endoscopy (SBCE) is increasingly used in coeliac disease (CD); however, its diagnostic performance in non-responsive (NRCD), refractory (RCD), and complicated coeliac disease (CCD) has not been systematically evaluated. Current guidelines recommend SBCE as a second-line investigation in suspected RCD or CCD, following serology, histology, and assessment of dietary adherence, yet no prior systematic review or meta-analysis has focused specifically on these high-risk cohorts. We introduce treatment-unresponsive coeliac disease (TUCD) as an umbrella term encompassing NRCD, RCD, and CCD. We aimed to determine the diagnostic yield (DY), clinical impact, and safety of SBCE in adults with TUCD.Methods We systematically searched MEDLINE, EMBASE, Cochrane Central, and PubMed from 1 January 2000 to 30 November 2025 for studies evaluating SBCE in CD patients with persistent symptoms or suspected refractory or complicated disease. Two reviewers independently performed screening and data extraction, assessed risk of bias using ROBINS-E, and evaluated certainty of evidence using GRADE. We performed random-effects meta-analyses using a generalised linear mixed model to estimate pooled DY for RCD and secondary outcomes (ulcerative lesions and small bowel (SB) tumours). Random-effects meta-regression fitted using restricted maximum likelihood, explored whether DY was associated with prespecified covariates, including mean age, ulceration DY, tumour DY, completion rate, coeliac serology, proportion with known RCD, and symptom burden.Results Seventeen studies involving 979 adults were eligible. The mean age was 50.9 years, and 28.7% were male. Median disease duration was approximately 7.1 years among studies with extractable data. Pooled DY for SBCE in TUCD was 82% (95% CI 67–91%), with subgroup yields of 79% for NRCD (95% CI 44–95%; I 2=64%) and 99% for RCD (95% CI 82–100%; I2=0%). SBCE led to management change in 57% of patients (95% CI 21–87%; I2=92%). Pooled DY was 3% for ulcerative jejunitis (95% CI 1–7%; I2=64%) and 3% for SB malignancy (95% CI 1–6%; I2=0%). Meta-regression demonstrated that the proportion of patients with known RCD and the proportion experiencing management change were significantly associated with DY. Capsule retention occurred in 0.5% (95% CI 0.02–1.4%); only one study reported patency capsule use. The conversion rate to device-assisted enteroscopy (DAE) was 18% (95% CI 9–30%).Conclusion SBCE has a high diagnostic yield in TUCD, particularly in the RCD subgroup, with a relatively low frequency of ulcerative jejunitis and SB malignancy. These findings support SBCE as a second-line, non-invasive investigation and a filter test before DAE in TUCD. Future studies should evaluate whether artificial intelligence improves detection and characterisation of TUCD on SBCE.