Adjustment of Surveillance Intervals for Ulcerative Colitis-Associated Neoplasia Based on Disease Duration

2.1 Patient Selection and Data Collection

This study was conducted as a retrospective observational investigation that included patients diagnosed with UC-associated high-grade dysplasia or adenocarcinoma at Keio University Hospital, Tokyo, Japan, between September 2010 and February 2023. As documented in prior studies [2, 3, 12-16], the included cases were diagnosed endoscopically and confirmed as UCAN by final histopathological evaluation performed by at least two experienced pathologists with expertise in UCAN. Patients with the following diagnoses without UCAN were excluded from the cohort: non-dysplastic lesions, neuroendocrine tumors, sporadic adenomas/carcinomas, serrated lesions, traditional serrated adenomas, and indefinite UCAN. Comprehensive demographic and clinical data were collected, including patient sex, age at UCAN detection and at one prior colonoscopy, duration of UC, disease extent, anatomical location of UCAN, and the presence of concomitant primary sclerosing cholangitis. Exclusion criteria were strictly applied to ensure cohort uniformity. Patients with a UC duration of < 8 years at the time of colonoscopy prior to UCAN diagnosis were excluded. Similarly, patients lacking complete colonoscopic imaging from the preceding examination in which UCAN was identified were not included. Furthermore, patients undergoing follow-up at intervals of < 1 year—either for disease monitoring or after the detection of low-grade dysplasia that eventually progressed to high-grade dysplasia or adenocarcinoma—were excluded, as such cases did not align with the study's objectives, which focused on standard surveillance practices.

2.2 Evaluation of Endoscopic and Histologic Findings

The diagnostic approach for UCAN, integrating histological assessment with advanced endoscopic techniques, such as indigo carmine dye spraying and magnifying endoscopy, has been previously reported in the literature [2, 16]. In the present study, UCAN cohort data diagnosed by these established methods were used. Detailed procedures are presented in Data S1.

2.3 Assumption of Lesion Identification Rate

The analysis included cases with available data from both the most recent colonoscopy in which UCAN was not diagnosed and the first colonoscopy in which UCAN was detected. The date of the former examination was designated as the date of the previous colonoscopy (D_pre), whereas the latter was defined as the date of UCAN detection (D_det). To evaluate whether lesions could be identified within the surveillance intervals recommended by various guidelines, the recommended date for the next colonoscopy (D_rec) was calculated on the basis of D_pre (Figure 1A). Patient risk stratification for UCAN was conducted according to the American Society for Gastrointestinal Endoscopy (ASGE) [4] guidelines, which classified risk into two levels, and the American Gastroenterological Association (AGA) [5], European Crohn's and Colitis Organisation (ECCO) [6], and British Society of Gastroenterology (BSG) [7] guidelines, each of which stratified risk into three levels (Table 1). The Japanese Society for Gastrointestinal Endoscopy [8] and the American College of Gastroenterology [9] guidelines, which do not specify intervals based on stratified risk levels, were excluded from this analysis. Across the included guidelines, common high-risk criteria included severe disease activity, complications of primary sclerosing cholangitis, bowel stenosis, and a family history of early-onset colorectal cancer. For high-risk cases meeting any of these criteria, D_rec was uniformly set as 1 year after D_pre. For guidelines specifying a surveillance interval range (e.g., 1–3 years or 2–3 years), D_rec was standardized to 3 years after D_pre to distinguish from the 1-year surveillance interval of the high-risk group. For guidelines with three risk levels, D_rec for the low-risk group, defined by the absence of risk factors, was set at 5 years after D_pre. If D_rec fell within 180 days of D_det, the endoscopic interval was classified as successful for lesion identification. In cases where D_rec preceded D_det by 1 year, lesion identification was also deemed successful, as a subsequent D_rec would occur in the following year. On the basis of these criteria, lesion identification rates were calculated for each guideline, accounting for their respective risk stratification frameworks and surveillance interval recommendations.

2.4 Risk-Adjusted Assumptions Based on Disease Duration

Three models were evaluated to assess alterations in lesion identification rates following adjustments for disease duration as a risk factor (Figure 1B). The duration of disease was categorized into four threshold patterns: ≥ 15, ≥ 20, ≥ 25, and ≥ 30 years. In Model 1, cases exceeding a specific disease duration were classified as intermediate risk, with a reassessment interval of 3 years. In Model 2, cases were classified as high risk, necessitating annual surveillance, a strategy deemed neither cost-effective nor practical. To address this limitation, Model 3 introduced a stepwise risk escalation, increasing the risk level by one tier (e.g., from low risk to intermediate risk or from intermediate risk to high risk). Each model incorporated risk adjustments based on disease duration, and corresponding changes in the D_rec were calculated. These calculations also accounted for the adjusted lesion identification rates under each model.

2.5 Statistical Analysis

Continuous variables were reported as medians with interquartile ranges. Categorical variables were summarized as percentages. Intergroup comparisons were made using Pearson's chi-squared test or Fisher's exact test. Comparisons between the original lesion identification rates for UCAN and the adjusted rates under each model were performed using the McNemar test, which was also applied to determine the statistical significance of differences in lesion identification rates across varying disease durations. Two-sided p values < 0.05 were considered statistically significant.

3.1 Patient Profile

A total of 64 patients who underwent colonoscopy at our institution between September 2010 and February 2023 were diagnosed with UC-associated high-grade dysplasia or adenocarcinoma. Among these, exclusions were applied to patients with a duration of UC < 8 years at D_pre (n = 8; 12.5%), those with a short interval between colonoscopies (n = 8), and those for whom complete endoscopic imaging data at D_pre were unavailable for review (n = 9). After these exclusions, a total of 39 patients were included in the analysis (Figure 2).

3.2 Patient Characteristics

The demographic and clinical characteristics of the 39 patients with UCAN are summarized in Tables 2, 3 and S1. The cohort consisted of 27 men (69%) and 12 women (31%), with a median age at UCAN diagnosis of 52 years (46–67 years). The median duration of UC was 22 years (16–30 years), with a median disease duration of 21 years (14–27 years) at D_pre. The median colonoscopy interval (D_det − D_pre) was 1.3 years (1.1–2.2 years). When stratified by lesion depth, no significant differences in colonoscopy interval were observed: Tis lesions (n = 19) had a median interval of 1.3 years (1.1–2.2 years); T1 lesions (n = 6), 1.5 years (1.2–1.8 years); T2 lesions (n = 4), 1.2 years (1.0–1.7 years); and lesions deeper than T3 (n = 10), 1.3 years (1.0–1.5 years). Of the included patients, six (15%) did not meet the surveillance interval criteria recommended by established guidelines. Among the 33 surveillance colonoscopies conducted within recommended intervals, 16 (48%) detected lesions with a depth greater than T1 (Figure 3). Eleven patients (28%) were in both endoscopic and histologic remission at the time of assessment.

3.3 Adjustment of Surveillance Intervals by Disease Duration Increases Lesion Identification Rate

Risk stratification based on the criteria outlined in each guideline demonstrated that the intermediate- and high-risk groups constituted the majority of cases (Table 4). A comparison of the D_rec determined by guideline-specific risk stratification, starting from D_pre to the actual D_det, revealed lesion identification rates of 72% for ASGE, 59% for AGA, 44% for ECCO, and 56% for BSG guidelines. These were used as baseline data and adjusted for the duration of disease (Figure 4).

Model 1 was initially applied, categorizing patients with disease durations of ≥ 15 years (n = 29), ≥ 20 (n = 20), ≥ 25 (n = 13), and ≥ 30 (n = 7) as intermediate risk. However, given the paucity of low-risk cases in the cohort, no significant changes in lesion identification rates were observed, precluding meaningful analysis (Table 4, Figure 4A). Subsequently, Model 2 was implemented, reclassifying patients with disease durations of ≥ 15, ≥ 20, ≥ 25, or ≥ 30 years as high risk. For instance, under the ASGE guideline, the lesion identification rate was initially 72% (28/39). Reclassification of eight patients with disease durations ≥ 15 years (7/29 identified lesions) led to an adjusted identification rate of 90% (35/39). Corresponding rates for durations of ≥ 20, ≥ 25, and ≥ 30 years were 87%, 82%, and 72%, respectively. Similarly, risk reassessments under other guidelines yielded increased identification rates, as follows: 90%, 87%, 82%, and 72% for ASGE; 85%, 82%, 74%, and 64% for AGA; 82%, 74%, 62%, and 49% for ECCO; and 82%, 79%, 72%, and 62% for BSG (Figures 1 and 4B, Table 4). These findings indicate that shortened surveillance intervals in patients with prolonged disease durations were associated with significantly improved lesion identification rates.

Using Model 3, which increased the risk category by one rank for disease durations of ≥ 15, ≥ 20, ≥ 25, or ≥ 30 years (Figures 1 and 3), lesion identification rates mirrored those of Model 2. Specifically, rates for ASGE were 90%, 87%, 82%, and 72%; for AGA, 85%, 82%, 74%, and 64%; for ECCO, 82%, 74%, 62%, and 49%; and for BSG, 82%, 79%, 72%, and 62%, respectively (Figure 4C). Because low-risk participants were almost absent in this cohort, the results for Models 2 and 3 were largely identical. Compared with pre-adjustment rates, significant differences in lesion identification were noted for ASGE (≥ 15 years: p = 0.023; ≥ 20 years: p = 0.041), AGA (≥ 15 years: p = 0.004; ≥ 20 years: p = 0.008; ≥ 25 years: p = 0.041), ECCO (≥ 15 years: p < 0.001; ≥ 20 years: p = 0.002; ≥ 25 years: p = 0.023), and BSG (≥ 15 years: p = 0.004; ≥ 20 years: p = 0.008; ≥ 25 years: p = 0.041). These results indicate that in our cohort, increasing risk stratification levels and shortening surveillance intervals for patients with disease durations of ≥ 20–25 years significantly enhanced lesion identification rates, regardless of the guideline applied.