Introduction

Detection of metachronous advanced colorectal neoplasia is the main purpose of surveillance colonoscopy. Current guidelines are largely based on observational stuides or expert opinions which may not accurately predict the chance of metachronous advanced neoplasia.

Aims & Methods

This study aimed to assess the performance of machine learning (ML) models to predict the presence of advanced neoplasia during surveillance colonoscopy and also attempted to decipher the factors associated with advanced neoplasia as determined by machine learning models. The training dataset consisted of 348 patients who had undergone surveillance colonoscopy between 2016 and 2019 in our hospital. Dataset included 20 baseline clinical parameters during the index colonoscopy, findings and performance metric of the index colonoscopy. Tweleve different advanced machine learning models were trained to predict the presence of advanced neoplasia during surveillance colonoscopy. Advanced neoplasia was defined as the presence of adenoma with high grade dysplasia, ≥1 cm adenoma or sessile serrated polyp (SSP) , ≥ 3 adenoma or SSP, SSP with dysplasia or traditional sessile adenoma (TSA). Trained models were further validated in another cohort of 320 patients with surveillance colonoscopy between 2020 and 2021. The performance was determined by the area under the receiver operating characteristic curve (AUC). The feature of importance was computed by Gini importance.

Results

Metachronous advanced neoplasia was present in 32.7% (n=114) and 24.4% (n=78) in the training and validation set, respectively. Among the 12 machine learing models, the Grandient Boosting Classifier (GBC) had the highest AUC (0.85, 95%CI: 0.83-0.86). The sensitivity, specificity, positive and negative predictive value of the GBC was 0.77, 0.77, 0.29 and 0.96, respectively. Miss rate was 3.7% with number need to screen of 3.4 in the validation set. The other 11 models have AUC ranging from 0.70 to 0.84, which were all significantly higher than the performance if follow the USMSTF (0.60) and ESGE (0.51) recommendations. Feature extraction of the GBC model showed that the time between index and surveillance colonoscopy (44.9%), patient’s age (15.5%), and withdrawal time (8.1%), adenoma detection rate (ADR) (7.7%) and number of polyp of index colonoscopy (6.6%) were the top five predictive factors for advanced neoplasia on surveillance colonoscopy (Table 1).

Table 1: Feature of importance of GBC model

Predictors	Feature of importance (%)
Time between surveillance colonoscopy and index colonoscopy	44.9
Patient’s age	15.5
Withdrawal time of index colonoscopy	8.1
Adenoma detection rate of index endoscopist	7.7
Number of polyp in index colonoscopy	6.6
Number of advanced adenoma in index colonoscopy	4.7
History of diabetes mellitus	3.0
Number of advanced serrated lesions in index colonoscopy	2.7
History of ischemic heart disease Patient’s sex	1.8 1.5

Conclusion

Most machine learning models could help to predict the chance of metachronous advanced neoplasia during surveillance colonoscopy and are better than current recommendations. Feature extraction shows that apart from interval between index and surveillance colonsocopy, the performance metrics and findings of index colonoscopy play pivotal role on detection of advanced neoplasia in surveillance colonoscopy.

Introduction

Despite the availability of various image-enhanced endoscopy (IEE) techniques, direct comparisons of their performance in detecting lower gastrointestinal lesions are scare.

Aims & Methods

This systematic review and network meta-analysis aims to compare the diagnostic performance of different IEE modalities on detection of colorectal lesions. We conducted a systematic search of PubMed, EMBASE, and Web of Science for randomized controlled trials (RCTs) published up to January 1, 2025, which compared the diagnostic accuracy of various IEE in detecting colorectal lesions. The primary outcomes were rate-based metrics including polyp (PDR), adenoma (ADR), advanced adenoma (AADR), and serrated lesion detection rate (SLDR). Additionally, count-based measures were compared, including mean polyp per patient (MPP) and mean adenoma per patient (MAP). Narrow-band imaging (NBI) techniques were classified into two categories: old generation (e.g., EVIS LUCERA Spectrum or EXERA II) and new generation (e.g., EVIS EXERA III or LUCERA ELITE). To rank the performance of each modality, we employed the Surface Under the Cumulative Ranking curve (SUCRA). For binary outcomes, odds ratios (ORs) were calculated; and for continuous outcomes, mean differences (Δ) were determined with all results reported alongside their respective 95% confidence intervals (CIs).

Results

IEE modalities(PDR)	SURCA	Probaility of being the best	Mean Rank
CE	92.6	55.8	1.6
TXI	79.2	34.7	2.7
LCI	60.5	0.8	4.2
NBI(New generation)	60.2	2.1	4.2
IEE modalities(ADR)	SURCA	Probability of being the best	Mean Rank
TXI	96.9	89.3	1.2
LCI	79.7	6.7	2.4
CE	55.4	1.5	4.1
NBI(New generation)	48	0.5	4.6

A total of 45 RCTs involving 55,165 participants were included. Linked color imaging (LCI) and new-generation NBI were the most frequently evaluated IEE modalities. For PDR, the highest SUCRA are seen with chromoendoscopy (CE, 92.6), texture and color enhancement imaging (TXI, 79.2), and LCI (60.5). For ADR, TXI showed the best performance (SUCRA: 95.9), followed by LCI (79.7) and CE (55.4). In advanced adenoma detection, TXI (76.4) and CE (69.4) performed best. Compared to white light imaging (WLI), LCI (Δ: 0.71, 95% CI: 0.55–0.92) and TXI (Δ: 0.58, 95% CI: 0.34–0.96) significantly increased PDR. CE was the only technique significantly superior to WLI in both ADR (Δ: 0.80, 95% CI: 0.64–0.99) and AADR (Δ: 0.66, 95% CI: 0.45–0.97). For continuous outcomes, LCI (SUCRA: 75.3) and new-generation NBI (74.7) performed best for MPP; and new-generation NBI (82.1) and TXI (80.2) ranked highest for MAP. Few studies reported detection rates of serrated lesion and no significant differences were found among IEE methods due to limited data.

Conclusion

This network meta-analysis demonstrates that TXI, new-generation NBI, LCI, and CE all exhibit better performance than WLI in detecting colorectal lesions. TXI demonstrated consistent efficacy in both polyp and adenoma detection. The new-generation NBI and LCI showed significant advantages in terms of lesion yield per patient. However, the diagnostic benefit of IEE for serrated lesions remains uncertain due to insufficient evidence. Additional head-to-head studies are necessary to further confirm these findings.

Introduction

The indications for colonoscopy are well established, with the primary objective being the detection of neoplasia or lesions that carry a high risk for colorectal cancer, even though benign causes are more commonly encountered. In young adults under 50 years of age, common indications include rectal bleeding and diarrhea. In contrast, individuals aged 50 years and older are screened more broadly due to the increased risk of colorectal cancer, although recent trends show a rising incidence of neoplastic lesions in younger adults as well. ​

Aims & Methods

The aim of our study is to compare the indications and outcomes of colonoscopy between individuals younger and older than 50 years, and to identify which indications are most strongly associated with neoplastic findings.
A retrospective, descriptive, and analytical study was conducted over a period of 3.5 years, involving 1,518 patients. Participants were divided into two groups: Group A included patients under 50 years old, while Group B comprised those aged 50 and above. Epidemiological, clinical, and endoscopic data were compared between the two groups. Patients with a known diagnosis of inflammatory bowel disease (IBD) were excluded from the study. Statistical analysis was carried out using SPSS version 21.0.

Results

Among the patients, 65.1% were over 50 years of age, while 34.9% were under 50. The mean age in group A (under 50) was 37.5 ± 9 years, compared to 64.4 ± 9 years in group B (over 50). Females predominated in the younger group (55.3%), whereas males were more prevalent among older patients (56.3%) (p < 0.001). A personal history of colorectal cancer was more frequent in patients over 50 (10.5% vs. 4.4%; p < 0.001).
Colonoscopy indications varied by age group: constipation was the most common indication in group B (29.1% vs. 22%; p = 0.01), whereas diarrhea predominated in group A (33.3% vs. 19.3%; p < 0.001). Pathological findings on colonoscopy were more frequent in older patients (49.1% in group B vs. 32.9% in group A; p < 0.001).
In patients under 50, the most common pathological findings were colitis or ulcerative colitis (47.9% vs. 11.7%; p < 0.001). In contrast, patients over 50 showed a higher prevalence of polyps (61.5% vs. 41.3%; p < 0.001) and rectocolonic neoplasia (16% vs. 5%; p = 0.002).
Among the colonoscopy indications, constipation in patients under 50 was associated with a higher risk of neoplastic lesions (p = 0.007), while in patients over 50, rectal bleeding and constipation were significantly associated with neoplastic lesions (p < 0.001).

Conclusion

The presence of neoplastic lesions was strongly associated with constipation in individuals under 50 years of age, and with both rectal bleeding and constipation in those over 50. Further studies to better define these indications could help improve the appropriateness of colonoscopy, particularly in younger patients.

Introduction

Effective withdrawal time (EWT), as measured by EndoScreener QC (Wision A.I.), is a novel artificial intelligence (AI)-derived colonoscopy quality metric, which measures the actual duration of clear images obtained during scope withdrawal from cecum to rectum. We have previously demonstrated the superiority of EWT to standard withdrawal time (SWT) on adenoma detection rate (ADR) in a retrospective study.(1) This prospective study determines the association between EWT and colonic lesion detection rates.

Aims & Methods

This was a prospective evaluation of the use of real-time determination of EWT, as measured by the EndoScreener QC, of endoscopists who were blinded to the output of the system. Patients were recruited in our hospital from Nov 2023 to Feb 2024. All examination was performed with high-definition endoscopes (EVIS-EXERA 290 video system, Olympus Optical, Tokyo, Japan) under white light. In all examination, colonoscope was first advanced to the cecum as confirmed by identification of the appendiceal orifice and ileocecal valve or by intubation of the ileum. After cecal intubation, the colonoscopy was slowly withdrawn. The AI derived (AI) real time effective mucosal examination monitor system (EndoScreen QC) was initiated during scope withdrawal, starting from cecum to anus. The SWT was defined as total withdrawal time from cecum to rectum after excluding polypectomy and biopsy time. All detected polyps were removed during scope withdrawal only. The size (measured with biopsy forceps), location and morphology of each polyp was recorded by an independent observer. The bowel preparation quality was graded according to the Boston Bowel Preparation Scale. The primary outcome was the correlation between EWT and ADR, which was determined by the Pearson correlation coefficient (r); and compared with SWT. Other outcomes included correlation between EWT and polyp detection rate (PDR) or serrated lesion detection rate (SDR).

Results

In the prospective analysis, 185 colonoscopies performed by 9 endoscopists were analyzed. The mean age was 67.4 year (SD: 10.5) and 57.3% (n=106) were male. 83.9% (n=154) were diagnostic colonoscopy. The median Boston Bowel Preparation score (BBPS) was 7. The mean ADR was 49.2% (95% CI: 41.6-56.2, n=91), mean EWT was 4.4 (95% CI: 4.1-4.6) minutes and mean SWT was 7.1 (95% CI: 6.7-7.5) minutes. The correlation between EWT and ADR was significantly higher (r=0.98, 95%CI: 0.96-1.00) than between SWT and ADR (r=0.80, 95%CI: 0.53-0.98, p<0.01). For PDR, the correlation between EWT and PDR was also significantly higher than between SWT and PDR (r=0.97, 95%CI: 0.95-1.00 vs r=0.73, 95%CI: 0.30-0.98, p=0.02). However, there was no significant difference between EWT and SWT on SDR [r=0.92, 95%CI: 0.72-1.00 vs r=0.93, 95%CI: 0.86-0.98, p=0.89].

Table 1: Baseline characteristics and outcomes of the study

Number Mean age in years (SD) Male sex, % (n)	185 67.4 (10.5) 106 (57.3)
Indications: Screening, % (n) Diagnostic, % (n) Bowel preparation in BBPS, median (range) Mean standard withdrawal time in mins (95%CI) Mean EWT in mins (95%CI)	31 (16.8) 154 (83.9) 7 (6-9) 7.1 (6.7-7.5) 4.4 (4.1-4.6)
ADR %(n) PDR %(n) SDR % (n) Mean number of adenoma (SD) Mean number of polyp (SD) Mean number of serrated lesion (SD)	49.2 (91) 58.9 (109) 20.0 (37) 1.1 (0.1) 1.6 (0.1) 0.4 (0.1)
r between SWT and ADR (95%CI) r between EWT and ADR (95%CI) r between SWT and PDR (95%CI) r between EWT and PDR (95%CI) r between SWT and SDR (95%CI) r between EWT and SDR (95%CI)	0.80 (0.53-0.98) 0.98 (0.96-1.00) 0.73 (0.30-0.98) 0.97 (0.95-1.00) 0.93 (0.86-0.98) 0.92 (0.72-1.00)

Conclusion

The novel AI-derived EWT demonstrated better correlation with ADR and PDR than SWT in this prospective study. Further studies should evaluate the role of applying real-time monitoring EWT on endoscopy training and quality improvement of endoscopists with low ADR.

References

1. Lui TKL, Ko MKL, Liu JJ, Xiao X, Leung WK. Artificial intelligence-assisted real-time monitoring of effective withdrawal time during colonoscopy: a novel quality marker of colonoscopy. Gastrointest Endosc. 2024;99(3):419-27.e6.

Disclosure

The following companies had given donation to our team for research.
Olympus, Fujifilm, Medronic, Wision AI, Insign Medical Co Ltd

Introduction

Serrated polyps (SPs) are precursors to 15–20% of colorectal cancers (CRCs). However, there are uncertainties regarding which SPs require surveillance and at what intervals, with recommendations adapted from those for adenomas in the absence of solid evidence.

Aims & Methods

Our aim was to assess which SP risk characteristics relate to a higher risk of metachronous CRC or advanced polyps. We systematically searched PubMed, EMBASE, and Cochrane for cohort, case-control studies, and clinical trials from inception to Dec 31, 2023, for CRC or advanced polyps [advanced adenoma (AA) or advanced SP] incidence at surveillance stratified by baseline SP size, dysplasia, location, and multiplicity. CRC and advanced polyp incidence per 1,000 person-years (p-y) were estimated. We performed a meta-analysis by calculating pooled relative risks (RR) using a random-effects model.

Results

5,903 studies were reviewed and 14 included, with 493,949 patients (mean age 59·5 years, 55% men). Mean follow-up was 4·9 years. CRC incidence per 1,000 p-y was 2·09 (95%CI 1·29–2·90) for advanced SP, 1·52 (0·78–2·25) for SP>10mm, 5·86 (2·16–9·56) for SP with dysplasia, 1·18 (0·77–1·60) for proximal SP, 0·52 (0·08–1·12) for >3SP, 0·50 (0·35–0·66) for non-advanced SP, and 0·44 (0·41–0·46) for normal colonoscopy. Metachronous CRC risk was higher in advanced SP vs non-advanced SP (RR 1·84, 95%CI 1·11–3·04), and vs normal colonoscopy (RR 2·92, 2·26–3·77); in SP>10mm vs <10mm (RR 2·61, 1·43–4·77), and vs normal colonoscopy (RR 3·52, 2·17–5·69); and in SP with dysplasia vs normal colonoscopy (RR 2·71, 2·00–3·67). No increase in CRC or advanced polyp risk was found in patients with proximal vs distal SP, nor in >3SP vs 1–2SP.

Conclusion

CRC risk is significantly higher in patients with baseline advanced SP after 4·9 years of follow-up, with risk magnitudes similar to those described for AA, supporting the current recommendation for 3-year surveillance in patients with advanced SP.

Disclosure

RJ has received research grants from MSD, and has participated as an advisor to MSD, Norgine, Alpha-Sigma, and GISupply. UL has participated in the advisory board (UniversalDx, Lean Medical, Vivante, Kohler Ventures), and as consultant (Medtronic, Clinical Genomics, Guardant Health, Freenome, ChekCap). MDR has no COI since start of 2020 (Advisor for Norgine in 2019). The rest of authors have nothing to disclose.

Introduction

Patients with type 2 diabetes mellitus (T2DM) are at increased risk of developing colorectal adenomas and cancer. Newer medications such as sodium-glucose cotransporter-2 (SGLT2) inhibitor and glucagon-like peptide-1 receptor agonists (GLP-1 RA) are increasingly used in DM patients with promising results on improving glycaemic control and other outcomes. While studies showed that the use of GLP-1 RA could affect intestinal transit time and bowel preparation quality, their effect on adenoma detection remains unclear.

Aims & Methods

This population-based retrospective cohort study included all adult T2DM patients who were prescribed with either GLP-1 RA or SGLT2 inhibitors for more than 180 days, and had undergone colonoscopy between 2018 and 2024 in Hong Kong. Anonymous patient’s data were retrieved from the territory-wide electronic medical system (CDARS). We excluded patients with history of previous colonoscopy or were on both medications. The primary outcome was proportion of patients with adenoma detected (ADR) during colonoscopy in GLP-1 versus SGLT2 group. GLP-1 RA use was treated as a time-varying variable to address immortal-time bias. Multivariable Cox regression model with adjusted hazard ratio (aHR) was used as primary analysis, which adjusted for patient’s age and sex, smoking and drinking status, body mass index, colonoscopy’s quality (bowel preparation, scope withdrawal time and indications), time-weighted mean HbA1c, concurrent illnesses (hypertension, stroke, ischemic heart disease, chronic renal disease and chronic liver disease), medication uses (dipeptidyl peptidase 4, insulin, metformin, pioglitazone, statin, sulfonylurea, proton pump inhibitor). Propensity-score matching (PSM) with 1:2 match was used as secondary analysis.

Results

There was a total of 12,789 DM patients who had colonoscopy during the study period (977 on GLP-1 RA and 11812 on SGLT2). Among them, 6192 (48.4%) patients were found to have adenomas during colonoscopy (GLP-1 38.3% and SGLT2 49.2%). In the time-dependent analysis, GLP-1 group was found to have a significantly lower ADR when compared to SGLT2 group (aHR: 0.40; 95% CI 0.36–0.45; P<0.01). In the PSM cohort (1:2 match), all covariates were well matched with absolute standardized mean difference < 0.1. PSM analysis showed a similar finding with aHR: 0.38 (95% CI 0.33–0.43; P<0.01; Table 1). There was a duration-response relationship with longer GLP-1 use duration linked to greater reduction in ADR (<1 year: aHR 0.71, 95% CI 0.60–0.84; 1–<2 years: aHR 0.51, 95% CI 0.43–0.61; ≥2 years: aHR 0.33, 95% CI 0.28–0.38; P-trend<0.01). Consistent lower ADRs were observed across different levels of bowel preparation quality (from BBPS 3 to 9) in GLP-1 group with HRs of 0.37–0.45 (p ≤ 0.01). Subgroup analysis showed similar results across most demographics, clinical factors, and comorbidities (HRs 0.34–0.44, p < 0.01), except in subgroups with short scope withdrawal time.

Table 1. Association between GLP-1 RA use with ADR

Primary outcome of interest— Detection of adenoma	ADR	Adjusted HR (95% CI)	P values
Main analysis (time-dependent analysis)* SGLT2-I users GLP-1 RA users	49.3% (5818/11812) 38.3% (374/977)	Ref 0.40(0.36-0.45)	<0.01
Secondary analysis (PSM analysis) SGLT2-I users GLP-1 RA users	46.7% (912/1954) 38.3% (374/977)	Ref 0.38(0.33-0.43)	<0.01

Conclusion

DM patients who use GLP-1 RA are found to have a lower rate of adenoma detected during colonoscopy than those who use SGLT-2 in this population-based study. The results remain similar even after stratified according to different levels of bowel preparation quality. These findings may suggest a potential protective role of GLP-1 receptor agonists against colorectal adenoma development, which warrant further prospective study to confirm.

Introduction

The indications for colonoscopy are clearly defined and its main objective is to look for neoplasia or lesions at high risk of colorectal neoplasia, even if benign lesions are much more frequent. Common indications for young adults < 50 years include rectal bleeding and diarrhea, while for subjects over 50 years, the indications may be larger given the risk of colorectal cancer, despite the recent increase in neoplastic lesions in young adults.

Aims & Methods

A retrospective descriptive and analytical study, including 1518 patients, was carried out over 3 and a half years. The patients were divided into 2 groups: group A for patients aged less than 50 years, and group B for patients aged greater than 50 years. We compared epidemiological, clinical, and endoscopic data between the two groups. Patients known to have inflammatory bowel disease (IBD) were excluded from our study. The statistical analysis was carried out by the JAMOVI software.
Our study aims to compare the indications and results of colonoscopy between those under and over 50 years old and to determine the indications associated with a risk of neoplastic lesion.

Results

65.1% of patients were over 50 years old, and 34.9% were under 50 years old. The average age was 37.5 ± 9 years in group A and 64.4 ± 9 years in group B. The female gender predominated among those under 50 years old (55.3%), while the male gender predominated among those over 50 years old (56.3%) (p<0.001).
History of colorectal cancer predominated among those over 50 years old at 10.5% (vs 4.4%; p<0.001). The indications for colonoscopy were dominated by constipation in patients in group B at 29.1% (vs 22%; p=0.01) and by diarrhea in patients in group A at 33.3% (vs 19. 3%; p<0.001).
Colonoscopy was pathological in 49.1% of patients in group B vs 32.9% of patients in group A (p<0.001). It was dominated by colitis or rectocolitis (47.9%) in subjects under 50 years old (vs 11.7%; p<0.001), by polyps (61.5% vs 41.3%; p<0.001) and colorectal neoplasia (16% vs 5%; p=0.002) in older patients.
A higher risk of neoplastic lesions was associated with rectal bleeding and constipation (p < 0.001) in subjects > 50 years old, and with constipation in subjects < 50 years old (p = 0.007).

Conclusion

The presence of neoplastic lesions was strongly associated with the presence of constipation in subjects < 50 years old and with rectal bleeding and constipation in subjects > 50 years old. Early screening with colonoscopy might be beneficial for young individuals with these higher-risk lesions of colorectal cancer.