Skip Navigation
Skip to contents

Clin Endosc : Clinical Endoscopy

OPEN ACCESS

Articles

Page Path
HOME > Clin Endosc > Volume 57(6); 2024 > Article
Editorial Understanding the discrepancy in the effectiveness of artificial intelligence-assisted colonoscopy: from randomized controlled trials to clinical reality
Jung Ho Bae,orcid
Clinical Endoscopy 2024;57(6):765-767.
DOI: https://doi.org/10.5946/ce.2024.226
Published online: November 25, 2024

Department of Gastroenterology, Healthcare System Gangnam Center, Seoul National University Hospital, Seoul, Korea

Correspondence: Jung Ho Bae Department of Gastroenterology, Healthcare System Gangnam Center, Seoul National University Hospital, 39FL Gangnam Finance Center, 152 Teheran-ro, Gangnam-gu, Seoul 06236, Korea E-mail: newsanapd@naver.com
• Received: August 28, 2024   • Revised: October 14, 2024   • Accepted: October 15, 2024

© 2024 Korean Society of Gastrointestinal Endoscopy

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (https://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

  • 129 Views
  • 5 Download
prev next
See the article "Assessing the potential of artificial intelligence to enhance colonoscopy adenoma detection in clinical practice: a prospective observational trial" on page 783.
Artificial intelligence (AI) has emerged rapidly as an innovative technology in various medical domains, including gastroenterology. Among its most promising applications is the enhancement of polyp detection during colonoscopy, which has garnered significant attention and enthusiasm within the gastroenterologist community.1,2 The GI Genius module (Medtronic), the first commercial computer-aided polyp detection (CADe) system in the United States, has demonstrated substantial efficacy across multiple randomized controlled trials (RCTs) that employed both parallel and tandem study designs, with a marked improvement in adenoma detection rates (ADRs) and a reduction in the adenoma miss rate.3 However, as AI transitions from research settings to real-world clinical practice, a critical question arises: can these technologies maintain the same level of effectiveness amidst the complexities of routine clinical workflows?
The study by Rønborg et al.4 provides a pragmatic evaluation of the GI Genius module’s performance within a clinical setting. This single-center, prospective observational study aimed to compare ADRs among patients undergoing colonoscopy with and without the assistance of the GI Genius module, according to the routine endoscopy workflow over a 3-month period. The findings of this study offer a nuanced perspective on the impact of AI on routine clinical practice. The CADe group exhibited a slight, although not statistically significant, increase in ADRs (34.7% vs. 30.5%). Multivariable analysis further substantiated these findings, indicating a relative risk (RR) of 1.12 (95% confidence interval [CI], 0.88–1.43), which did not reach statistical significance. Despite the limitations of the small sample size of the study, these results suggest that while the GI Genius module may not offer sufficient benefit for adenoma detection in real practice, the overall effect appears to be modest at best.
The findings of this study align with the emerging evidence from other real-world clinical settings, indicating that the advantages of AI-assisted colonoscopy in routine practice may be less pronounced than those reported in RCTs. A recent meta-analysis on the use CADe systems in real-world practice, including 6 prospective and 6 retrospective studies, similarly demonstrated that the improvement in the ADR with CADe was marginal compared to standard colonoscopy (36.3% vs. 35.8%; RR, 1.13; 95% CI, 1.01–1.28).5 Notably, this observed benefit of CADe was absent in the subgroup analysis of the 6 retrospective studies (35.7% vs. 36.2%). Furthermore, in 4 studies utilizing the GI Genius module, no significant difference in the ADR was observed with CADe (RR, 0.96; 95% CI, 0.85–1.07).
The discrepancy between RCTs and real-world studies raises significant questions regarding the translation of AI-assisted colonoscopy from research settings to clinical practice. It is crucial to recognize several key differences between the study environments in RCTs and real-world studies. RCTs meticulously control various factors that influence outcomes, including patient selection and procedural standardization, such as withdrawal time, bowel preparation, and operator expertise. In contrast, real-world studies reflect the inherent complexities and variability of routine clinical practice. Procedures in real-world settings are conducted by a diverse group of endoscopists with varying skill levels and often involve suboptimal bowel preparation.
Moreover, the lack of blinding among endoscopists to the CADe intervention could introduce performance bias. This bias may have led to changes in behavior, such as a more diligent mucosal inspection or a more thorough lesion evaluation during AI-assisted colonoscopy. Awareness of CADe use may exert a stronger influence on performance outcomes in RCTs where conditions are more controlled. Although real-world studies cannot entirely eliminate this bias, they are likely to be less affected by such unconscious biases owing to the higher workload and time constraints inherent in routine clinical practice. Consequently, the true impact of CADe may be more accurately reflected in real-world settings, where endoscopists operate under less-controlled conditions.
One of the key challenges in implementing AI within clinical practice is the interaction between AI systems and endoscopists.6 User fatigue resulting from frequent false positives (FP) can lead to decreased satisfaction and reduced trust in critical alerts, potentially resulting in reduced clinical effectiveness.7,8 This effect may be particularly pronounced in real-world settings with high procedural volumes. Studies have shown that while many commercial models demonstrate similar sensitivity for polyp detection, there is considerable variability in FP rates depending on the specific product and version used.9 Moreover, a lack of understanding or excessive enthusiasm towards AI technology can lead to over-reliance on these systems, causing endoscopists to neglect their inspections. This over-reliance may paradoxically result in poorer adenoma detection outcomes, highlighting the need for a balanced integration of AI tools with human expertise.
The marginal gains in ADRs observed in the study by Rønborg et al.4 should not detract from the potential of AI in colonoscopy. Instead, these findings encourage a more tempered and realistic approach to the integration of AI in clinical practice. Future research should focus on large-scale pragmatic trials that reflect the diversity and complexity of real-world clinical environments. Additionally, there is a need to develop AI systems that are more adaptable to the variability of clinical practice, including tools that better account for the nuances of procedural contexts.
  • 1. Rey JF. As how artificial intelligence is revolutionizing endoscopy. Clin Endosc 2024;57:302–308.ArticlePubMedPMCPDF
  • 2. Sivananthan A, Nazarian S, Ayaru L, et al. Does computer-aided diagnostic endoscopy improve the detection of commonly missed polyps? A meta-analysis. Clin Endosc 2022;55:355–364.ArticlePubMedPMCPDF
  • 3. Lou S, Du F, Song W, et al. Artificial intelligence for colorectal neoplasia detection during colonoscopy: a systematic review and meta-analysis of randomized clinical trials. EClinicalMedicine 2023;66:102341.ArticlePubMedPMC
  • 4. Rønborg SN, Ujjal S, Kroijer R, Ploug M. Assessing the potential of artificial intelligence to enhance colonoscopy adenoma detection in clinical practice: a prospective observational trial. Clin Endosc 2024;57:783–789.Article
  • 5. Wei MT, Fay S, Yung D, et al. Artificial intelligence-assisted colonoscopy in real-world clinical practice: a systematic review and meta-analysis. Clin Transl Gastroenterol 2024;15:e00671.ArticlePubMedPMC
  • 6. Lee J, Cho WS, Kim BS, et al. Impact of user's background knowledge and polyp characteristics in colonoscopy with computer-aided detection. Gut Liver 2024;18:857–866.ArticlePubMedPMC
  • 7. Nehme F, Coronel E, Barringer DA, et al. Performance and attitudes toward real-time computer-aided polyp detection during colonoscopy in a large tertiary referral center in the United States. Gastrointest Endosc 2023;98:100–109.ArticlePubMed
  • 8. Tiankanon K, Karuehardsuwan J, Aniwan S, et al. Performance comparison between two computer-aided detection colonoscopy models by trainees using different false positive thresholds: a cross-sectional study in Thailand. Clin Endosc 2024;57:217–225.ArticlePubMedPMCPDF
  • 9. Troya J, Sudarevic B, Krenzer A, et al. Direct comparison of multiple computer-aided polyp detection systems. Endoscopy 2024;56:63–69.ArticlePubMedPMC

Figure & Data

REFERENCES

    Citations

    Citations to this article as recorded by  

      • PubReader PubReader
      • ePub LinkePub Link
      • Cite
        CITE
        export Copy Download
        Close
        Download Citation
        Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

        Format:
        • RIS — For EndNote, ProCite, RefWorks, and most other reference management software
        • BibTeX — For JabRef, BibDesk, and other BibTeX-specific software
        Include:
        • Citation for the content below
        Understanding the discrepancy in the effectiveness of artificial intelligence-assisted colonoscopy: from randomized controlled trials to clinical reality
        Clin Endosc. 2024;57(6):765-767.   Published online November 25, 2024
        Close
      • XML DownloadXML Download
      Related articles
      Understanding the discrepancy in the effectiveness of artificial intelligence-assisted colonoscopy: from randomized controlled trials to clinical reality
      Understanding the discrepancy in the effectiveness of artificial intelligence-assisted colonoscopy: from randomized controlled trials to clinical reality

      Clin Endosc : Clinical Endoscopy Twitter Facebook
      Close layer
      TOP