Endoscopy 2022; 54(12): 1180-1181
DOI: 10.1055/a-1890-5043
Editorial

Will artificial intelligence-assisted colonoscopy work in cancer screening programs?

Referring to Rondonotti E et al. p. 1171–1179
1   Digestive Disease Center, Showa University Northern Yokohama Hospital, Yokohama, Japan
› Author Affiliations

Fecal immunochemical testing (FIT) and subsequent colonoscopy are components of cancer screening programs worldwide, the aim being to reduce colorectal cancer (CRC) incidence and mortality. However, FIT-based screening schemes do not prevent all cancers: some patients who have undergone FIT-based colonoscopies develop post-colonoscopy colorectal cancer (PCCRC). One of the main causes of PCCRC is poor-quality colonoscopy. The adenoma detection rate (ADR) is an established quality indicator for primary screening colonoscopy [1], this rate being inversely associated with the incidence of PCCRC. Whether ADR is an equivalent indicator of quality in post-FIT colonoscopies and primary screening colonoscopies has not yet been established. However, it is generally believed that ADR serves as a reliable quality indicator in post-FIT colonoscopies [2].

With recent advances in artificial intelligence (AI) technology, computer-aided detection (CADe) systems for polyp detection during real-time colonoscopy have been attracting attention both in the academic field and marketplace. Generally, such CADe systems are reported to increase the ADR [3] and reduce the adenoma miss rate [4]. However, because most of these studies did not involve cancer screening programs, the clinical impact of CADe in such programs remains unknown.

“These findings imply that CADe can work well in the context of FIT-based screening programs regardless of the endoscopists’ baseline ADRs.”

In this issue of Endoscopy, Rondonotti et al. provide intriguing insights regarding the efficacy of CADe in a FIT-based screening program [5]. The authors conducted a multicenter, randomized controlled trial in five endoscopy centers in Italy. People with positive FIT results in a CRC screening program were recruited. Eligible individuals were randomly allocated to a CADe-assisted colonoscopy or standard high definition white-light (HDWL) colonoscopy arm. In the CADe group, endoscopists used a convolutional, neural network-based CADe system (CAD EYE; Fujifilm Co., Tokyo, Japan) to support their examination. All participating endoscopists were well qualified to participate in the FIT-based screening program, having achieved ADRs of at least 25 %. The study cohort comprised 800 patients (CADe arm 405; HDWL arm 395). ADR, a primary outcome of this study, was significantly higher in the CADe arm than in the HDWL arm (53.6 % [95 %CI 48.6–58.5] vs. 45.3 % [95 %CI 40.3–50.4]). The authors also calculated the ADR in the two study arms according to the endoscopists’ baseline ADR. Interestingly, the ADR was greater in the CADe arm regardless of the baseline ADR; however, this difference was not significant. In contrast, similarly to previous studies [3], an increment effect occurred only with diminutive (≤ 5 mm) polyps; there was no significant increase in the detection rate of advanced adenomas. However, there was insufficient statistical power to conclude that CADe does not improve the ADR for advanced adenomas.

These findings imply that CADe can work well in the context of FIT-based screening programs regardless of the endoscopists’ baseline ADRs. Although there is little evidence that ADR correlates with the incidence of PCCRC among FIT-positive individuals, it is possible that CADe will prove to reduce CRC incidence and mortality. Furthermore, a recent simulation study showed that using CADe in screening colonoscopies saves medical costs by reducing cancer incidence and mortality [6]. It will be very interesting to determine whether such cost-effectiveness occurs with application of AI in FIT-based screening programs. However, in the Rondonotti et al. study, CADe did not improve the ADR for advanced, small (6–9 mm), or large (< 10 mm) adenomas. Additionally, the types of lesions that are responsible for PCCRCs have not yet been accurately determined. Sessile serrated lesions [7] or nonpolypoid laterally spreading tumors [8] are said to be responsible for PCCRCs. Of note, CADe systems have not been proven to increase the detection rate of these lesions. Therefore, large prospective studies with lengthy follow-ups are needed to determine the direct cancer prevention effect of AI systems. One such large study is currently ongoing in EU countries [9].

In summary, Rondonotti et al. have conducted an interesting study that shows clinical benefits of using CADe in the framework of a FIT-based CRC screening program. CADe systems are already available in clinical practice. Despite the puzzle still missing some pieces, the era of AI in colonoscopy may have already begun.



Publication History

Article published online:
29 July 2022

© 2022. Thieme. All rights reserved.

Georg Thieme Verlag KG
Rüdigerstraße 14, 70469 Stuttgart, Germany

 
  • References

  • 1 Kaminski MF, Regula J, Kraszewska E. et al. Quality indicators for colonoscopy and the risk of interval cancer. N Engl J Med 2010; 362: 1795-1803
  • 2 Cubiella J, Castells A, Andreu M. et al. Correlation between adenoma detection rate in colonoscopy- and fecal immunochemical testing-based colorectal cancer screening programs. United European Gastroenterol J 2017; 5: 255-260
  • 3 Barua I, Vinsard DG, Jodal HC. et al. Artificial intelligence for polyp detection during colonoscopy: a systematic review and meta-analysis. Endoscopy 2021; 53: 277-284
  • 4 Kamba S, Tamai N, Saitoh I. et al. Reducing adenoma miss rate of colonoscopy assisted by artificial intelligence: a multicenter randomized controlled trial. J Gastroenterol 2021; 56: 746-757
  • 5 Rondonotti E, Di Paolo D, Rosa Rizzotto E. et al. Efficacy of a computer-aided detection system in a fecal immunochemical test-based organized colorectal cancer screening program: a randomized controlled trial (AIFIT study). Endoscopy 2022; 54: 1171-1179
  • 6 Areia M, Mori Y, Correale L. et al. Cost-effectiveness of artificial intelligence for screening colonoscopy: a modelling study. Lancet Digit Health 2022; 4: e436-e444
  • 7 Rex DK, Ahnen DJ, Baron JA. et al. Serrated lesions of the colorectum: review and recommendations from an expert panel. Am J Gastroenterol 2012; 107: 1315-1329 quiz 1314, 1330
  • 8 Matsuda T, Fujii T, Sano Y. et al. Randomised comparison of postpolypectomy surveillance intervals following a two-round baseline colonoscopy: the Japan Polyp Study Workgroup. Gut 2020; 70: 1469-1478
  • 9 UMIN-ICDR Clinical Trial. Artificial Intelligence in Colonoscopy for Cancer Prevention – a Randomized Health Service Implementation Trial. UMIN identification: UMIN000044748. Accessed: 29 June 2022: https://center6.umin.ac.jp/cgi-open-bin/icdr_e/ctr_view.cgi?recptno=R000051108