Artificial Intelligence cannot Transform Observational Non-difference into Clinical Equivalence
Endoscopy International Open · 23.09.2026 · 14 (CP)
DOI: 10.1055/a-2961-0699 ↗Kurz gefasst
Leserbrief in der Zeitschrift Endoscopy International Open, erschienen am 23.09.2026. Verfasst von Gabriele Wurm Johansson und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of GastroenterologySkåne University Hospital, Lund UniversityMalmöSweden. Die Arbeit nennt 6 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
To the Editor,Nevarez et al. should be commended for applying automated assessment to the unresolved problem of small-bowel (SB) cleanliness. 1 However, their conclusion that a liquid diet provides "adequate" cleanliness comparable to laxatives-and, therefore, questions routine purgative use-is not supported by the design or analysis.The central problem is allocation according to the usual practice of 11 participating hospitals rather than randomisation.The article does not show the centre-by-preparation distribution or account for centre clustering.If centres predominantly used different regimens, preparation may be inseparable from centre, including differences in case mix, instructions, referral pathways, and reading practices.The baseline imbalances in age and diabetes, both subsequently associated with cleanliness, confirm that the groups were not exchangeable.Moreover, p = 0.415 does not demonstrate comparability.The study was designed only to detect differences; it did not define in advance how close the results had to be for the "treatments" were genuinely equivalent.Failure to reject a difference is not evidence of equivalence.The problem is compounded because the sample-size calculation concerned a 20% difference in the proportion of "adequate" cleansing, whereas the primary analysis used a continuous Endoclean score.The stated power calculation, therefore, does not justify the analysis performed.The authors also acknowledge that no validated Endoclean cutoff defines adequate preparation.Nevertheless, "adequate" cleanliness becomes the clinical conclusion.Image-level accuracy of a convolutional neural network does not automatically validate the clinical meaning 1 Endoscopy International
Autor:innen und Institutionen
- Department of GastroenterologySkåne University Hospital, Lund UniversityMalmöSweden.
- Department of Clinical ResearchUniversity of Southern DenmarkSvendborgDenmark.
- Department of SurgeryOdense University Hospital and Svendborg HospitalSvendborgDenmark.
- Department of GastroenterologyPomeranian Medical UniversitySzczecinPoland.
ORCID-Profile: Anastasios Koulaouzidis, Ervin Toth
Zitieren
APA
Wurm Johansson, G., Koulaouzidis, A., & Toth, E. (2026). Artificial Intelligence cannot Transform Observational Non-difference into Clinical Equivalence. Endoscopy International Open, 14(CP). https://doi.org/10.1055/a-2961-0699
Vancouver
Wurm Johansson G, Koulaouzidis A, Toth E. Artificial Intelligence cannot Transform Observational Non-difference into Clinical Equivalence. Endoscopy International Open. 2026;14(CP). doi:10.1055/a-2961-0699
BibTeX
@article{WurmJohansson2026,
author = {Wurm Johansson, Gabriele and Koulaouzidis, Anastasios and Toth, Ervin},
title = {Artificial Intelligence cannot Transform Observational Non-difference into Clinical Equivalence},
journal = {Endoscopy International Open},
year = {2026},
volume = {14},
number = {CP},
doi = {10.1055/a-2961-0699}
}
EUVIMED