Original Title: Kunstig intelligens som beslutningsstøtte ved koloskopiske undersøgelser til diagnosticering af neoplastisk sygdom

Authors' objectives: In recent years, several manufacturers of colonoscopes have begun promoting the use of AI-assisted colonoscopies, including the use of computer-aided detection (CADe). While some studies highlight the positive effects of AI in improving adenoma detection rates (ADR), others express the concerns of overdiagnosis. In this health technology assessment, the Danish Health Technology Council examined the current evidence of the use of AI as clinical decision-support in colonoscopy for diagnosis of neoplastic disease.

Authors' results and conclusions: Clinical effectiveness and safety: The results of two RCTs indicate that the use of AI in colonoscopy is associated with only a clinically relevant difference in ADR. The expert committee expects the increase in ADR to be attributed to a statistically significant increased detection of adenomas < 10 mm, which are likely to be low-risk findings, and therefore of no clinical relevancy. Based on current evidence, there is no data to inform whether the use of AI has any positive effects on overall survival rate, quality of life or complication rates. The overall assessment of the expert committee is that CAD-assisted colonoscopy is equivalent to standard colonoscopy. Patient perspective: Based on seven qualitative studies of patient perspectives on the general use of clinical AI three themes where identified, concerning attitudes and acceptance, benefits and apprehensions, and the patient-clinician-AI relationship. From these themes a number of patient- and non-patient-related factors that influence people's attitudes towards clinical AI were identified, as well as patient perceived benefits and apprehensions. The expert committee concludes that there are no patient concerns that speak for or against the use of clinical AI. Organizational implications: Based on an interview study, a thematic analysis was carried out, identifying five themes concerning purpose and implementation, increased detection, workload, quality indicators in colonoscopy, and future potentials. The expert committee assesses that it is possible to establish uniform national implementation of the technology, but that the expected increased ADR is unlikely to be commensurate with the investment. Health economics: The results of a cost-effectiveness analysis (CEA) and budget impact analysis (BIA) indicate that CAD-assisted colonoscopy is associated with an increase in costs compared to standard colonoscopy attributed to a higher technology cost and increased rates of polyp resection. The expert committee concludes that there are no health-economic implications for or against the application of CADe-assisted colonoscopy due to a lack of health-economic evidence supporting CADe-assisted colonoscopy across indications and the use of patient-related outcomes.

Project Status: Completed

URL for project: https://behandlingsraadet.dk/find-evalueringer-analyser/analyser/2021/kunstig-intelligens-som-beslutningsstotte-ved-koloskopiske-undersogelser-til-diagnosticering-af-neoplastisk-sygdom

Year Published: 2023

URL for published report: https://behandlingsraadet-classic.azureedge.net/media/bbjjjro3/use-of-artificial-intelligence-as-clinical-decision-support-in-colonoscopy.pdf

English language abstract: An English language summary is available

Publication Type: Full HTA

Country: Denmark

Organisation Name: The Danish Health Technology Council

Contact Address: Niels Jernes Vej 6a, 9220 Aalborg

Contact Name: Nikolaj Hellmuth Skak

Contact Email: nsp@behandlingsraadet.dk