Inter and intra-observer agreement for the LST Classification in Large (>2cm) Colorectal Laterally Spreading Tumours
Roupen Djinbachian, Jérémie Jacques, Victoire Michal, Ludovico Alfarone, Robert Bechara, Nicholas G Burgess, Mariana Figueiredo, Yusuke Fujiyoshi, Lucile Heroin, Michal F. Kaminski, Eric Lam, Philippe Leclercq, Isabelle Lienhart-Chambon, Alexandru Lupu, Charles Menard, Jeffrey Mosko, Nima Mottacki, Heiko Pohl, Douglas K. Rex, Enrique Rodriguez De Santiago, Yutaka Saito, Amrita Sethi, Sandro Sferazza, Neal Shahidi, Daniel von Renteln, Dennis Yang, Mathieu PiocheBackground:
Morphological assessment of large (≥20 mm) colorectal laterally spreading tumors (LSTs) is central to predicting submucosal invasion risk, procedural difficulty, and optimal resection strategy. The LST classification is widely used; however, its diagnostic reliability has never been formally evaluated.
Patients and Methods:
We performed a prospective multicenter video-based study. 24 blinded expert endoscopists from North America, Europe, Asia, and Oceania independently classified video recorded LSTs twice, in randomized order, into 5 prespecified LST subtypes. Inter- and intra-observer agreement were calculated using Light’s kappa. Secondary analyses evaluated agreement by LST size category, endoscopist sex, geographic training region, and a simplified 3-category granular vs non-granular classification.
Results:
46 LSTs were evaluated, yielding 1104 classifications. Inter-observer agreement for the primary outcome was poor in both the first (κ=0.37) and second assessment rounds (κ=0.36). Agreement varied modestly across LST size groups and endoscopist characteristics, with moderate agreement in Europe and Oceania and fair agreement in North America and Asia. Simplifying morphology into 3 categories improved inter-observer agreement to the moderate range (κ≈0.48). Intra-observer agreement ranged widely (κ=0.28–0.94), with the highest reproducibility among endoscopists trained in Europe and Oceania. Intra-observer agreement was generally higher for smaller lesions (<40 mm) and improved when using the 3-category model.
Conclusions:
Among international experts, inter-observer agreement for the LST classification is low, with considerable inter- and intra-observer variability. These findings highlight the need for improved standardization and suggest caution when using detailed LST subclassification for clinical decision-making or as an endpoint in research.