Article type
Abstract
Background: The integration of Computer Aided Detection (CADe) systems, an artificial intelligence (AI) tool, into colonoscopy signifies a major advancement in AI application, potentially enhancing detection of polyps and patient outcomes. Despite being the most researched area for application of AI in medicine with rapid adoption, there are ongoing challenges in ensuring safe and ethical use while minimizing bias with no clear standards for evaluation of AI technologies. Here, we report on a living guideline initiative, namely the BMJ Rapid Recommendations on CADe in adults undergoing colonoscopy, performed as part of a European Union-funded grant (OperA).
Methods: An international, multidisciplinary panel (21 individuals), including patient partners, gastroenterologists, primary care providers, and methodologists, prioritized the questions and patient-important outcomes. The evidence to support the guideline was (1) a recently published systematic review (SR) of 22 randomized trials on CADe, (2) a commissioned microsimulation study to assess the long-term benefits and harms, and (3) a SR on values and preferences. Using the GRADE framework, we summarized findings, rated our certainty in conclusions, and developed recommendations. For the patient perspective, the panel prioritized the balance between benefits and harms, overall certainty in conclusions, and patient values and preferences, whereas for the societal perspective, they also considered resource utilization, equity, acceptability, and feasibility.
Results: We suggest against the use of CADe for adults undergoing colonoscopy. Low certainty evidence suggests a trade-off between increased patient burdens (more surveillance colonoscopies) and small benefits (reduced colorectal cancer incidence and mortality) and variability on patient values and preferences and acceptability of adopting CADe by care providers. Incorporating the societal perspective, there was anticipated increased resource utilization which could impact equitable access to colonoscopy, as well as anticipated variable acceptability of adopting CADe by care providers.
Conclusion: Despite its promise, CADe's current application may increase healthcare burdens without clear evidence of significant benefit, necessitating further research and dynamic guideline updates to ensure its effective and ethical use in colorectal cancer screening and diagnosis. Our comprehensive approach highlights the feasibility of applying the standards and methods of evidence-based medicine and trustworthy, living guidelines to the rapidly evolving field of AI.
Methods: An international, multidisciplinary panel (21 individuals), including patient partners, gastroenterologists, primary care providers, and methodologists, prioritized the questions and patient-important outcomes. The evidence to support the guideline was (1) a recently published systematic review (SR) of 22 randomized trials on CADe, (2) a commissioned microsimulation study to assess the long-term benefits and harms, and (3) a SR on values and preferences. Using the GRADE framework, we summarized findings, rated our certainty in conclusions, and developed recommendations. For the patient perspective, the panel prioritized the balance between benefits and harms, overall certainty in conclusions, and patient values and preferences, whereas for the societal perspective, they also considered resource utilization, equity, acceptability, and feasibility.
Results: We suggest against the use of CADe for adults undergoing colonoscopy. Low certainty evidence suggests a trade-off between increased patient burdens (more surveillance colonoscopies) and small benefits (reduced colorectal cancer incidence and mortality) and variability on patient values and preferences and acceptability of adopting CADe by care providers. Incorporating the societal perspective, there was anticipated increased resource utilization which could impact equitable access to colonoscopy, as well as anticipated variable acceptability of adopting CADe by care providers.
Conclusion: Despite its promise, CADe's current application may increase healthcare burdens without clear evidence of significant benefit, necessitating further research and dynamic guideline updates to ensure its effective and ethical use in colorectal cancer screening and diagnosis. Our comprehensive approach highlights the feasibility of applying the standards and methods of evidence-based medicine and trustworthy, living guidelines to the rapidly evolving field of AI.