Comparative Diagnostic Accuracy of Cholangioscopy-Based Modalities for Indeterminate Biliary Strictures: A Systematic Review and Network Meta-Analysis
摘要
Indeterminate biliary strictures present a diagnostic challenge with a differential ranging from benign conditions such as IgG4-related disease to cholangiocarcinoma. ERCP-guided brush cytology has been the long-standing tissue-sampling strategy, but its reported sensitivity for malignancy is limited, frequently falling below 50% in published series. Complementary and alternative approaches have been developed, including fluorescence in situ hybridization (FISH) of biliary brushings, intraductal biopsy (IDB), and single-operator cholangioscopy (SOC) with the SpyGlass DS platform. These modalities have been evaluated individually, but a unified comparative synthesis has not previously been undertaken. This network meta-analysis (NMA) was conducted to synthesize the available evidence across these five strategies and to provide comparative rankings of their diagnostic performance.
MethodsWe searched PubMed, EMBASE, the Cochrane Library, and Scopus from inception through 31 December 2024, without language restriction. Eligible studies enrolled adult patients with indeterminate biliary strictures and evaluated at least one diagnostic modality against a histopathological or follow-up reference standard. Reporting followed methodological guidance for network meta-analyses of diagnostic test accuracy, with relevant items from the PRISMA-NMA and PRISMA-DTA extensions used to structure the review. A Bayesian hierarchical random-effects model was used to pool diagnostic accuracy data. Primary endpoints were sensitivity and specificity for malignancy. Secondary endpoints included diagnostic odds ratio (DOR), area under the ROC curve (AUROC), tissue adequacy, and adverse events. Surface under the cumulative ranking curve (SUCRA) values were calculated to rank modalities. Heterogeneity, network consistency, publication bias, and pre-specified sensitivity analyses (including restriction to SpyGlass DS-era data and to studies with a uniform reference standard) were formally assessed.
ResultsThirty-eight studies enrolling 4,912 unique patients met our criteria, contributing 6,570 per-modality observations across five diagnostic strategies: brush cytology, FISH, intraductal biopsy, SOC (SpyGlass DS), and combined FISH + brush cytology. SOC achieved the highest pooled sensitivity at 82% (95% CrI: 76%–88%; SUCRA 92%), and brush cytology had the lowest at 45% (95% CrI: 38%–52%; SUCRA 30%). The pattern was inverted for specificity: brush cytology was highest at 97% (95% CrI: 94%–99%) and SOC was lowest at 89% (95% CrI: 84%–93%). For overall diagnostic accuracy, SOC ranked first (SUCRA 88%), followed by FISH + brush cytology (SUCRA 80%). Pairwise comparison showed SOC was statistically superior in sensitivity to brush cytology, FISH alone, and intraductal biopsy; the comparison between SOC and FISH + brush cytology did not reach statistical significance (OR 1.43; 95% CrI: 0.97–2.13). SOC was associated with a higher adverse-event rate than cytology-based sampling (6.8 vs. 2.1%; OR 2.3, 95% CrI: 1.2–4.6); no procedure-related deaths were reported. Network consistency was acceptable, and no material publication bias was detected.
ConclusionsSOC (SpyGlass DS) demonstrated the highest pooled sensitivity for malignancy in indeterminate biliary strictures, but its interpretation should be guided by clinical context rather than statistical ranking alone. The sensitivity difference between SOC (SpyGlass DS) and combined FISH + brush cytology was not statistically significant in pairwise comparison, supporting FISH + brush cytology as a high-value, lower-morbidity alternative where SOC is unavailable or procedural risk is a concern. SOC was associated with a higher adverse-event rate than cytology-based sampling and showed lower specificity, which carries clinical implications for false-positive diagnoses. Choice of modality should therefore be considered in the context of local diagnostic capacity, patient-level pretest probability, and the relative consequences of false-negative and false-positive results. Comparative cost-effectiveness and the performance of these modalities in community-practice settings warrant further study.