Delivering a high-quality colonoscopy service fit for the 21st century

Abstract

Colorectal cancer (CRC) is the third most prevalent cancer globally. There is a concerning increase in its incidence among younger individuals. Colonoscopy remains the gold standard for CRC diagnosis. With the introduction of population-based bowel screening and increased public awareness, there has been a significant rise in referrals for colonoscopy. Healthcare providers worldwide will need to strategically evaluate how to allocate resources to adequately train the next generation of colonoscopists who will need to provide accurate endoscopic assessment and treatment for premalignant polyps and early CRC. This review outlines the current workload challenges faced by colonoscopists whilst exploring emerging technologies such as artificial intelligence for adenoma detection. Additionally, advanced endoscopic surgical techniques like endoscopic submucosal dissection are discussed.

Key Words: Colonoscopy; Colorectal cancer; Bowel screening; Polyps; Adenoma

Core Tip: This comprehensive review summarizes the current state of colonoscopy worldwide. We discuss workload, polyp detection and endoscopic management, training, and emerging technologies.

INTRODUCTION

Colorectal cancer (CRC) is the third most common cancer worldwide, and the second leading cause of cancer-related deaths internationally. Worldwide in 2022, more than 1.9 million new cases were diagnosed, and more than 900000 cancer-related deaths were recorded[1]. There is evidence that the incidence of both colon and rectal cancers are increasing in individuals younger than 50 years[2]. When these types of cancer are diagnosed early, survival and quality of life are significantly improved. Stage I colon and rectal cancers have a 5-year survival rate > 90%, whereas the 5-year survival rates for stage IV colon and rectal cancers are 13% and 17%, respectively[3].

The majority of CRCs develop through the ‘classical’ adenoma-carcinoma sequence, a process that can take 10 years or longer. CRC that developed via the relatively accelerated microsatellite instability pathway takes a few years[4]. The long latent phase in disease development means CRC is a potentially preventable and treatable disease.

Globally, over 50 million colonoscopies were performed annually in 2018 and 2019[5]. Colonoscopy involves direct high-resolution visualization of the colonic and rectal mucosa. It provides accurate detection and assessment of premalignant adenomas and carcinomas. In addition, advanced endoscopic resection techniques allow selected premalignant and early malignant lesions to be excised with good oncological outcomes without the need for conventional resection surgery. As a result, global demand for colonoscopy is expected to increase, which will put pressure on healthcare systems to provide this technology. This review summarized contemporary scientific literature to quantify the impact of colonoscopy in reducing disease burden, to evaluate the current quality indicators that define a good colonoscopy examination, to detail current standards in the assessment and endoscopic treatment of premalignant polyps, and to evaluate the current role of artificial intelligence in colonoscopy.

METHODS AND RESOURCES

A list of subtopics related to colonoscopy was compiled with consensus between the authors. A systematic literature search of PubMed, Medline, ScienceDirect, Cochrane Central Register of Controlled Trials databases, and Google Scholar was carried out by the authors. Keywords used included “colorectal cancer,” “colonoscopy,” “adenoma,” “polypectomy,” “quality,” “training,” and “artificial intelligence.” A summary of the literature included in the review is presented in Table 1.

Table 1 Summary of literature reviewed.

Ref.	Study format	Level of evidence1
Benefits and risks of colonoscopy
Bretthauer et al[6], 2022	Randomized controlled trial	I
Lin et al[7], 2021	Systematic review	II
Ryhlander et al[6], 2022	Cross-sectional study	III
Sajid et al[9], 2015	Systematic review	II
Hafner et al[10], 2015	Systematic review	II
Quality assurance, training, and human resources
Rex et al[11], 2015	Systematic review	II
Sawhney et al[12], 2008	Cross-sectional study	III
Chung et al[13], 2010	Cross-sectional study	III
Haycock et al[14], 2010	Randomized controlled trial	II
Khan et al[15], 2012	Systematic review	II
Barton et al[16], 2012	Cross-sectional study	III
Sedlack et al[17], 2011	Observational study	V
Gupta et al[18], 2012	Observational study	V
Hitchins et al[19], 2018	Review	V
Endoscopic assessment and risk stratification of premalignant polyps, with tattoo
Kaltenbach et al[20], 2022	Guideline	V
The Paris Endoscopic Classification of Superficial Neoplastic Lesions, 2003	Retrospective case series	V
Komeda et al[22], 2017	Retrospective case series	V
Pohl et al[23], 2018	Commentary	V
Kudo et al[24], 1996	Case series	V
Hayashi et al[25], 2013	Case series	V
Hewett et al[26], 2012	Case series	V
Kitajima et al[27], 2004	Case series	V
Ishigaki et al[28], 2020	Retrospective case series	V
Nivatvongs et al[29], 1991	Retrospective case series	V
Ueno et al[30], 2004	Retrospective case series	V
Cooper et al[31], 1995	Retrospective case series	V
Haggitt et al[32], 1985	Retrospective case series	V
Kikuchi et al[33], 1995	Case series	V
Heafner et al[34], 2014	Review	V
Yang et al[35], 2017	Review	V
Artificial intelligence
Repici et al[36], 2022	Randomized controlled trial	I
Hassan et al[38], 2020	Retrospective case series	V
Bowel preparation
Lai et al[39], 2009	Case series	V
Mahmood et al[40], 2018	Systematic review	II
Russmann et al[41], 2007	Retrospective cohort study	IV
Parikh et al[42], 2011	Review	V
Spada et al[43], 2017	Randomized controlled trial	I
Chen et al[44], 2020	Meta-analysis	II
Alternatives to colonoscopy
Lin et al[7], 2021	Systematic review	II
Rex et al[45], 2015	Case series	V

¹Adopted from “Level of evidence hierarchy” published by Elsevier Author Services.

BENEFITS AND RISKS OF COLONOSCOPY

When colonoscopy is used as a cancer screening tool in the asymptomatic population the risk of CRC is decreased by 18% compared to no screening at 10 years of follow-up [0.98% vs 1.20%; relative risk (RR): 0.82; 95% confidence interval (CI): 0.70-0.93][6]. In the intention-to-treat analysis, the risk of CRC-related death was lower in the screening group (0.84% vs 1.2%; RR: 0.69; 95%CI: 0.55-0.83). In the adjusted per-protocol analysis, the magnitude of reduction in CRC-related death was lower in the screening group (0.15% vs 0.30%; RR: 0.50; 95%CI: 0.27-0.77). The magnitude of risk reduction appeared to be smaller compared to previous studies[7]. This finding likely reflects a declining risk of CRC and an improvement in the prognosis of CRC due to better treatment options.

In the pooled analysis, the sensitivities of colonoscopy for the detection of adenomas ≥ 10 mm and ≥ 6 mm were 89%-95% and 75%-93%, respectively. The respective specificities are 89% and 95% respectively[7]. Although computed tomography (CT) colonography offers similar performance levels, colonoscopy has several benefits for the management of premalignant and malignant colorectal lesions including direct visualization and characterization of lesion morphology, tissue diagnosis, and the potential to resect the lesion. In addition, CT colonography has several disadvantages including less accurate assessment of distal rectal lesions and detection of ‘incidentomas,’ which are of no significant clinical consequence but require additional resources for follow-up and can generate unnecessary anxiety for the patient. The risks of colonoscopy are bleeding and perforation, which are < 18 per 10000 and < 6 per 10000, respectively[7].

The potential for pain during colonoscopy is a disadvantage. The endoscopist’s ability to assess pain during the procedure is just as important as performing the test itself. In one study examining patients with moderate/severe pain during colonoscopy, factors associated with underestimation of moderate/severe pain by endoscopists included being male [odds ratio (OR): 1.79], normal body mass index (OR: 1.71), no previous abdominal surgery (OR: 1.81), and first colonoscopy (OR: 1.81)[8]. The choice of insufflation technique is also a determining factor. Both carbon dioxide and water insufflation are shown to be superior compared to air insufflation. In one systematic review, use of carbon dioxide insufflation was associated with lower procedural pain [standardized mean difference (SMD): -1.34; 95%CI: -2.23 to -0.45; Z = 2.96; P < 0.003)] and postprocedural pain at 1 h (SMD: -1.11; 95%CI: -1.83 to -0.38; Z = 2.97; P < 0.003), and at 6 h and 24 h (OR: 0.44; 95%CI: 0.23-0.85; Z = 2.44; P < 0.01). Carbon dioxide insufflation was associated with faster intubation to the cecum (SMD: -0.20; 95%CI: -0.37 to -0.02; Z = 2.23; P < 0.03), but the cecal intubation rate was similar (P = 0.59) in both colonic insufflation techniques[9]. In another systematic review, water infusion subjects experienced significantly lower pain (mean difference in pain score on a 0 to 10 scale: -1.57; 95%CI: -2.00 to -1.14; P < 0.00001). Fewer participants requested on-demand sedation or analgesia or both (RR: 1.20; 95%CI: 1.14-1.27; P < 0.00001)[10].

QUALITY ASSURANCE, TRAINING, AND HUMAN RESOURCES

Adenoma detection rate (ADR), cecal intubation rate, and withdrawal time are all commonly used as quality indicators. ADR was initially set at > 20% (25% in males and 15% in females) based on earlier studies but was subsequently raised to 25% by the United States Task Force[11]. The minimal scope withdrawal time was set at 6 min, based on correlation with ADR. The withdrawal time of > 7 min did not improve ADR[12].

Basic competency in performing colonoscopy is achieved by colonoscopists after approximately 200 procedures, although this number varies between 100-500 procedures[13]. Simulation is now widely used to improve training for novices in conjunction with supervised hands-on training. Simulator training significantly improved performance in simulated cases compared to patient-based training. Time to completion and technical finesse were both improved in the simulator group, although such outcomes were not observed when the study groups were assessed in live cases[14].

A Cochrane review in 2012 showed virtual reality endoscopy is useful as an effective supplement to early conventional endoscopy training (apprenticeship model)[15]. However, there was no consensus in the literature for integration into an existing training program, for the duration of training, nor for the level of supervision. The Mayo Colonoscopy Skills Assessment Tool and the Direct Observation of Procedural Skills (DOPS) are used to assess skills. The Mayo Colonoscopy Skills Assessment Tool is generally used for trainees, whereas DOPS is generally used for more senior endoscopists[16,17]. A separate polypectomy-specific DOPS has also been validated[18]. Non-technical skill domains, such as situational awareness, decision-making, communication, teamwork, and leadership, have been developed but have not been validated. Such skills can potentially improve performance and patient experience[19].

ENDOSCOPIC ASSESSMENT AND RISK STRATIFICATION OF PREMALIGNANT POLYPS

As the quantity and quality of colonoscopies increase, the volume of premalignant and malignant polyp detection will also increase. All colonoscopists should be able to identify and characterize these polyps at the index colonoscopy. The United States Multisociety Task Force on Colorectal Cancer has set out a benchmarking guideline for lesion assessment and description and endoscopic removal[20]. As a minimum, the location, size, and morphology of the polyp and its surface should be carefully documented using the Paris Classifications[21,22]. All polyps should be photographed at the index procedure prior to removal. In particular, all lesions ≥ 10 mm in size should have the post-resection mucosal defect documented photographically. The use of electronic- (e.g., narrow band imaging) or dye (chromoendoscopy)-based image-enhanced endoscopy techniques may help improve morphological descriptions.

All trained colonoscopists should be able to resect small mucosal lesions (< 9 mm) using the cold or hot snare technique for complete resection. The cold and hot snare techniques are acceptable for non-pedunculated (10-19 mm) lesions by colonoscopists with sufficient polypectomy experience. However, endoscopic resection of polyps ≥ 20 mm should be delegated to endoscopists with advanced training. Endoscopic mucosal resection (EMR) and endoscopic submucosal dissection (ESD) are both acceptable techniques[23]. ESD has the advantage over EMR due to the en bloc nature of resection, which leads to better lateral and deep margins and more accurate pathological staging. However, the learning curve for ESD is steep, and the perforation rate is higher than in EMR. General anesthesia (and therefore hospitalization) is often needed due to the prolonged nature of the procedure. Therefore, the availability of ESD is limited to selected tertiary referral centers. Even if the index colonoscopy was performed by a colonoscopist with advanced training, EMR and ESD should be deferred to obtain informed consent of the risks.

Recognition of deep submucosal invasion in flat, depressed, or sessile lesions, regardless of polyp size, is key to determining between endoscopic resection and surgical resection. Key endoscopic appearance includes general ulceration of the lesion surface, disruption of the normal vascular, and pit patterns[24]. These findings correspond to type III by the NICE classification and type V/Vn by the Kudo classification[25,26]. Surface ulceration and stiffness of the lesion and colonic wall are also predictive of deep submucosal invasion. In the presence of such a lesion, the risk of lymph node metastasis is higher. Therefore, surgical resection is more suitable for good surgical candidates[27]. Pedunculated polyps can be resected with snare polypectomy with as much stalk as possible for the margin.

Lateral spread tumors (LSTs) are premalignant lesions > 10 mm with an extending horizontal growth pattern[28]. Non-granular LSTs have more potential for deep submucosal invasion and multifocal invasion compared to granular LSTs. Superficial submucosal invasion cannot be reliably identified after piecemeal resection. For lesions > 20 mm, ESD is the only modality that can achieve en bloc resection. Therefore, recognition of this subtype and referral to the appropriate multidisciplinary team to address a complex polyp is recommended.

When submucosal invasion (i.e. cancer) is identified in a resected polyp specimen, an assessment of the probability of residual cancer in the bowel wall and lymph node involvement is made. For pedunculated polyps, the presence of one of the unfavorable histological criteria warrants surgery in those who are good candidates. These criteria include poor differentiation[29], tumor budding (defined as separate small groups of cancer cells at the leading edge of the cancer)[30], and < 1 mm resection margin[31]. Haggitt level 4 (cancer invading the polyp stalk) is an alternative factor indicating a poor prognosis[32].

For non-pedunculated polyps, the presence of poor tumor differentiation, lymphovascular invasion, and tumor budding are considered poor histological criteria. A non-pedunculated malignant polyp is generally considered a higher risk for residual disease following endoscopic resection than a pedunculated malignant polyp. However, if the non-pedunculated polyp is resected en bloc and appropriately handled by the colonoscopist and pathologist, low-risk non-pedunculated polyps can be managed without surgery. Historically, the depth of invasion in a non-pedunculated polyp is measured by the Kikuchi Classification[33], which subdivides the mucosa into thirds. However, as full-thickness resection is not achieved for colonic endoscopic resections, the Katajima criterion of greater or less than 1000 µm of invasion has become the new standard in Japan[29].

For malignant rectal lesions, consideration should be made to perform full-thickness resection using the transanal approach by a trained colorectal surgeon. A baseline endoscopic ultrasound should also be performed. The risk of local recurrence at any stage is higher for rectal cancer compared to colon cancer[34]. Therefore, more intensive endoscopic surveillance is required if endoscopic resection is chosen for low-risk lesions.

It is imperative that lesions ≥ 10 mm or lesions that have high-risk endoscopic features are tattooed at the time of the index colonoscopy (regardless of their resection status). Currently, there is no consensus for the tattoo placement in relation to the polyp/polypectomy site, for the requirement of multiple spots, and for tattooing rectal lesions[35]. In our clinic, we tattoo using Spot^® (GI Supply, Camp Hill, PA, United States), which consists of highly purified carbon particles as a tattooing material in an attempt to reduce the mild tissue inflammation caused by traditional India ink. A saline submucosal bleb is created 2 cm distal to the lesion/polypectomy site before injection of the tattoo. At least two spots on opposite sites are injected to easily identify the polyp if the patient proceeds to surgery. Whichever technique is used, it is important that the colonoscopist documents where the tattoo is in relation to the polyp/polypectomy site to aid future follow-up.

ARTIFICIAL INTELLIGENCE

Artificial intelligence, specifically computer-aided detection (CADe) software, has been extensively studied in the context of improving ADRs. In a prospective, randomized controlled non-inferiority trial, CADe (GI Genius™ Intelligent Endoscopy Module; Medtronic, Minneapolis, MN, United States) improved ADR (53.3% vs 44.5%; RR: 1.22; 95%CI: 1.04 to 1.40; P < 0.01) in non-expert endoscopists (< 2000 lifetime colonoscopies)[36]. Similar results were achieved in another prospective multicenter randomized controlled trial where the overall ADR (39.9% vs 32.4%; P < 0.001), advanced ADR (6.6% vs 4.9%; P = 0.041), ADR of expert (42.3% vs 32.8%; P < 0.001) and nonexpert endoscopists (37.5% vs 32.1%; P = 0.023), and adenomas per colonoscopy (0.59 ± 0.97 vs 0.45 ± 0.81; P < 0.001) were all significantly higher in the artificial intelligence-assisted colonoscopy (Eagle-Eye; Xiamen Innovision Medical Technology, Xiamen, China)[37].

Deep learning, a subtype of machine learning, uses an artificial neural network comprised of layers of interconnected computing units that mimic biological neural connections and allow for complex interpretation of input data. This allows the computer program to learn from unstructured input data. CADe aids in the detection of small polyps otherwise missed by the human eye. However, further research and development are needed before CADe can be entrusted with full polyp characterization. Currently, there is a high false alarm rate (as high as 27.3 per colonoscopy)[38]. Initially, CADe is being trained to confirm hyperplastic polyps in the rectosigmoid area, which would reduce morbidity and costs for unnecessary polypectomies.

BOWEL PREPARATION

Poor bowel preparation adversely affects completion rate and ADR, and increases the cost due to a repeat examination. The Boston bowel preparation scale has been shown to be a valid and reliable measure of bowel preparation[39]. The quality of bowel preparation is affected by the type of bowel preparation, split-dose regimen, low-fiber diet, comorbidities, concomitant medications, inpatient status, and elderly age[40]. Polyethylene glycol (PEG) is commonly used in 2 L or 4 L preparations. It is safe in patients with renal impairment, although patients with heart disease should exercise caution due to the potential increase in plasma volume[41,42]. Split-dose high-volume PEG has been shown to be more effective than split-dose low-volume PEG. However, high-volume preparations often decrease compliance. As a result, an adjunct to PEG has been added. When 2 L of PEG plus citrate was compared to 4 L of PEG a similar cleansing efficacy (73.6% vs 72.3%; 95%CI: -7.5 to 10.1) with greater tolerability (25.4% vs 37.0%; P < 0.01) and acceptability (93.9% vs 82.2%; P < 0.001) were observed[43].

A low residue or clear liquid diet at least 1 d prior to colonoscopy are both acceptable. A low residue diet was shown in one study to have better tolerability (OR: 1.92; 95%CI: 1.36-2.70; P < 0.01). Patients in the same study had a higher willingness to repeat the bowel preparation (OR: 1.86; 95%CI: 1.34-2.59; P < 0.01)[44]. In patients who continue to have inadequate preparation despite compliance, a clear liquid diet for 2 d is usually prescribed, followed by a morning procedure. Other options include the addition of a second laxative if no contraindications exist (e.g., 4 L of PEG followed by 1 L of magnesium citrate) or repeating the preparation over 2 d (with the exception of sodium phosphate).

ALTERNATIVES TO COLONOSCOPY

CT colonoscopy and capsule colonoscopy are both comparable in diagnostic yield to colonoscopy. The sensitivity for adenomas ≥ 6 mm is 86% for CT colonography and 88% for capsule colonoscopy[7,45]. Whilst both modalities are acceptable for the initial screening of CRC and adenomas, they do not offer therapeutic options. In particular, a CT colonoscopy can result in extracolonic findings that require further follow-up (up to 26.9%)[7]. The interpretation of these images also consumes a significant amount of manpower. The role of colonoscopy as a cost-effective one-stop option to screen and treat premalignant lesions in the same setting remains undisputed.

CONCLUSION

Worldwide demand for colonoscopy is increasing. With a high workload volume, an opportunity exists to streamline and standardize colonoscopy techniques and reporting. In particular, patients with malignant polyps should be stratified by risk to effectively manage treatment through colonoscopy or surgery. To accomplish this, colonoscopists should focus on improving diagnostic yield and quality. Colonoscopists should also participate in multidisciplinary meetings for CRC and counsel patients on the benefits of endoscopic resection. This effectively creates an alternative pillar in the management of CRC, in addition to surgery and chemoradiotherapy.

This article provided a broad and current overview of colonoscopy. It was not intended to be a systematic review but was intended to set the direction for further qualitative and quantitative studies of the subtopics in this growing field.