In the last several years, there has been increased emphasis on both evidence and value-based care, especially within gastroenterology. This paradigm shift is multifactorial and has been tied to both clinical outcomes as well as reimbursement. In 2015, consensus guidelines from the American Society of Gastrointestinal Endoscopy (ASGE) and the American College of Gastroenterology (ACG) were published, highlighting key quality indicators within gastroenterology.1 Adenoma detection rate (ADR) was one of the primary quality indicators highlighted, with rates of ≥25% in patients aged >50 years and undergoing screening colonoscopy (20% in women and 30% in men) recommended.1 In addition to ADR ≥25%, a mean withdrawal time of ≥6 minutes was also recommended.1  

ADR and withdrawal time have subsequently become a routine part of documentation during colonoscopy, along with submission of this information to the GI Quality Improvement Consortium. Increased scrutiny of ADRs has led to multiple studies evaluating miss rates during colonoscopy. A systematic review of 6 studies including 465 patients was conducted by Jeroen C. van Rijn, MD, PhD, of the Department of Clinical Epidemiology and Biostatistics at the University of Amsterdam, and his colleagues. They found a pooled adenoma miss rate of 22% (95% CI, 19%-26%).2 Not surprisingly, smaller adenomas had higher miss rates; the miss rate for adenomas measuring 1 mm to 5 mm was 26% (95% CI, 27%-35%) compared with 13% for adenomas measuring 5 mm to 10 in size (95% CI, 8%-18%). Similarly, a meta-analysis of 43 publications conducted by Shengbing Zhao, MD, of the Department of Gastroenterology at Changhai Hospital in Shanghai, China, and colleagues found a 26% miss rate for adenomas (95% CI, 23%-30%), with proximal advanced adenomas (14%; 95% CI, 5%-26%) and flat adenomas (34%; 95% CI, 24%-45%) representing specific subtypes that posed a significant problem.3  

One of the main reasons that ADRs are strongly emphasized in the 2015 joint ASGE and ACG guidelines is the link to clinical outcomes.1 Multiple studies have shown that patients who undergo endoscopies with lower ADRs have increased risk of interval colorectal cancer (CRC).4,5 ADR is inversely associated with risk of interval CRC, advanced-stage interval cancer, and fatal interval cancer.5 A study conducted by Douglas A. Corley, MD, PhD, of the Division of Research at Kaiser Permanente in Oakland, California, and colleagues and published in the New England Journal of Medicine found that the risk for interval cancer decreased 3% for every 1% increase in ADR.5 

Based on the importance of ADR as a quality indicator, there has been increased research interest in developing modalities to improve ADRs. One example with accruing clinical data is the use of artificial intelligence (AI) via a computer-aided polyp detection (CADe) system.6 The AI program is “trained” to identify adenomas using thousands of videos of histologically confirmed polyps from multiple patients. During a colonoscopy, the AI program will provide real-time analysis of the mucosa and provide a “bounding box” around potential adenomas.6  


Continue Reading

Recently, Alessandro Repici, MD, of the Digestive Endoscopy Unit at the Humanitas Research Hospital in Rozzano, Italy, and colleagues conducted the Artificial Intelligence for Colorectal Adenoma Detection Rate (AID) study (ClinicalTrials.gov Identifier: NCT04079487) to evaluate the safety and efficacy of a new CADe system in the detection of colorectal neoplasia and published the findings in Gastroenterology.6 This randomized, multicenter trial was conducted across 3 Italian endoscopy centers. Patients aged 40 to 80 years (61.32 ± 10.2) underwent colonoscopy for either CRC screening or polyp surveillance. The study also included those patients with positive fecal immunochemical tests (FITs). Patients with signs or symptoms of CRC were also included; however, there was a lack of detailed information regarding these patients in the study.

The study included 685 patients (337 men) randomly assigned to undergo high-definition (HD) colonoscopy with or without CADe.6 The ADR was significantly higher in the CADe group (54.8%) compared with the control group (40.4%; relative risk, 1.30; 95% CI, 1.14-1.45). In addition, the CADe group had more adenomas per colonoscopy (mean 1.07±1.54) compared with the control group (mean 0.71±1.20; incidence rate ratio, 1.46; 95% CI, 1.15-1.86). Adenomas <10 mm were detected in significantly higher proportions in the CADe group compared with the control group. The CADe group also had higher proportions of detection of proximal and distal adenomas.6

There were no statistically significant differences in withdrawal time, detection of adenomas ≥10 mm, resection of non-neoplastic lesions, proportion of patients with at least 1 sessile serrated lesion, and detection rate of advanced adenomas on histology between the groups.6 The study authors concluded that utilization of real-time CADe with HD colonoscopy resulted in a 30% increase in ADR and a 46% increase in adenomas per colonoscopy.6  

As noted by Dr Repici and colleagues, small adenomas can be easily missed.6 Even when identified, it can be difficult to distinguish adenomas from hyperplastic polyps or even normal mucosal tissue. In addition, there can be frequent discrepancies between the endoscopic appearance of diminutive (≤3 mm) polyps removed when histopathologic review is performed. 

Another study published in Endoscopy reviewed 644 colorectal lesions ≤3 mm in size that were diagnosed as adenomatous by the endoscopist and found that 15.4% of these lesions were given a diagnosis of normal mucosa following pathologic assessment.7 This discrepancy can be multifactorial and includes endoscopist error in missing the lesion when conducting the polypectomy with forceps or a snare and/or processing errors. However, it does highlight an area of improvement where AI may play an important role.

One potential way to help minimize the discrepancies between endoscopic and pathologic evaluation of diminutive polyps includes another AI clinical decision support solution. A study published in Gastroenterology evaluated 644 colorectal lesions ≤3 mm that were removed using cold forceps or cold snare polypectomy.8 The authors used an AI clinical decision support solution that had been validated in prior studies in order to potentially clarify discrepancies between endoscopic and pathologic diagnosis of colorectal lesions ≤3 mm. 

Of the total lesions evaluated, 71.1% had a concordant pathologic diagnosis, while the remaining 28.9% lesions had a discrepancy between endoscopic and pathologic diagnosis. These discrepancies included pathologic diagnosis of normal mucosa (15.4%), hyperplastic polyp (13.2%), and sessile serrated polyp (0.3%). The AI clinical decision support solution agreed with the endoscopic diagnosis in 89.6% of lesions. In those lesions that were discordant between endoscopy and pathology findings, the AI clinical decision support solution agreed with the endoscopic diagnosis in 90.3% of lesions.8 

The study authors concluded that a pathologic diagnosis of “normal mucosa” should be questioned if there is strong endoscopic evidence that the lesion is actually an adenoma, especially if using AI.8 Polyps of this size may appear to be of low clinical significance; however, changing a diagnosis from normal mucosa to an adenoma can have a significant impact on a patient’s future colonoscopy recall schedule. These findings also highlight the importance of taking high-quality pictures for documentation purposes during the procedure, especially if there is a discrepancy in the pathologic diagnosis. 

The initial research of using AI in CRC screening is promising and will only continue to develop in the future. Many of the current publications evaluating AI in CRC screening are from outside of the United States, leaving many United States-based gastroenterologists without experience of these modalities. Some of the challenges associated with routine use of AI in CRC screening include the availability of specialized sites with the necessary technology, costs and insurance coverage, and the need for expertise in developing the specific AI programs. The medicolegal implications will also have to be carefully weighed. Outside of polyp detection, it will also be interesting to see if the role of AI continues to expand within other areas of gastroenterology including detection of gastrointestinal bleeding, identification of tumors within the gastrointestinal tract, and detection of Barrett esophagus.9

Follow @Gastro_Advisor

References

1.     Rex DK, Schoenfeld PS, Cohen J, et al. Quality indicators for colonoscopy. Gastrointest Endosc. 2015;81(1):31-53.

2.     van Rijn JC, Reitsma JB, Stoker J, Bossuyt PM, van Deventer SJ, Dekker E. Polyp miss rate determined by tandem colonoscopy: a systematic review. Am J Gastroenterol. 2006;101(2):343-350.

3.     Zhao S, Wang S, Pan P, et al. Magnitude, risk factors, and factors associated with adenoma miss rate of tandem colonoscopy: a systematic review and meta-analysis. Gastroenterology. 2019;156(6):1661-1674.e11.

4.     Kaminski MF, Regula J, Kraszewska E, et al. Quality indicators for colonoscopy and the risk of interval cancer. N Engl J Med.  2010;362(19):1795-1803.

5.     Corley DA, Jensen CD, Marks AR, et al. Adenoma detection rate and risk of colorectal cancer and death. N Engl J Med. 2014 3;370(14):1298-1306.

6.     Repici A, Badalamenti M, Maselli R, et al. Efficacy of real-time computer-aided detection of colorectal neoplasia in a randomized trial. Gastroenterology. Published online May 1, 2020. doi:10.1053/j.gastro.2020.04.062

7.     Ponugoti P, Rastogi A, Kaltenbach T, et al. Disagreement between high confidence endoscopic adenoma prediction and histopathological diagnosis in colonic lesions < 3 mm in size. Endoscopy. 2019;51(3):221-226. 

8.     Shahidi N, Rex DK, Kaltenbach T, et al. Use of endoscopic impression, artificial intelligence, and pathologist interpretation to resolve discrepancies between endoscopy and pathology analyses of diminutive colorectal polyps. Gastroenterology. 2020;158(3):783-785. 

9.     Le Berre C, Sandborn WJ, Aridhi S, et al. Application of artificial intelligence to gastroenterology and hepatology. Gastroenterology. 2020;158(1):76-94.e2