Review Reports - Diagnostic Utility of Endoscopic Features and Endoscopic Ultrasonography for Ulcerative Colitis-Associated Neoplasia: A Retrospective Study on the Role of Endoscopic Submucosal Dissection as a Total Biopsy

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Results (Abstr.) and Conclusion (Disc.) give high accuracy levels like in a study with confirmatory design. However, this is a retrospective, single-center observational cohort study (04/2016 – 01/2025) with exploratory design. Overall, this study concludes that “meticulous endoscopic evaluation combined with EUS showed high diagnostic accuracy for distinguishing UCAN and assessing invasion depth”, which has not been convincingly proven. Undoubtedly, ESD en bloc is the best strategy for UCAN (“total biopsy” for diagnosis, compare ref. #4), when UCAN is not “out-of indication for ESD” (i.e. deep T1b or T2 – requiring surgical resection).

IBD Guidelines and experts do not agree on validity of Kudo pit pattern and JNET mNBI classifications for diagnosis of UCAN. But so far, the two classifications have not been rigorously analyzed for optical diagnosis of UCAN. However, this manuscript does not provide such an analysis nor indication criteria for UCAN-ESD due to the following shortcomings. Major Points:

Diagnostic histopathologic criteria for the primary objective (UCAN vs sporadic Ad.) are neither described in detail nor referenced in Methods (e.g. as it is given in ref. #10).
Different numbers (8 vs 15) are reported in Results for histopathologic UCAN diagnosis among 23 lesions (d ≥2 cm) for ESD in 1 vs Tab. 2 (= ? Optical diagnosis of UCAN).
No information is given, what proportion of patients (or neoplastic lesions) was studied in clinical/mucosal remission of UC. Ref.s #14 and #15 refer to diagnosis of such endoscopic severity of UC, but are only misquoted for “AI tailored to UCAN diagnosis” at the end of Discussion. Clinical remission is required for optical diagnosis of UCAN (compare ref. #10).
UCAN were only diagnosed in 23 lesions (d ≥2 cm) after ESD, but not at all in 189 lesions (d < 2cm; i.e. 141 sporadic neoplasias & 48 non-neoplastic lesions). Was there a difference in histopathologic work-up/analysis for specimens after ESD vs. other endoscopic resection? Did any of the 189 lesions not undergo strict histopathologic analysis (i.e. lesions for exclusion from UCAN analysis? Compare ref. #10). In ref. #6, 16 UCAN showed a diameter in the range of 1.2 – 8 cm.
Criteria for optical diagnosis of UCAN sporadic adenoma are not described in Methods: magnification (x-fold?) used for JNET and pit pattern grading, chromoendoscopy for pit pattern (types II – IV) with indigo carmine, types (V_i, V_n) with crystal violet. Please include few endoscopic images of UCAN and its margin with JNET and pit pattern V_i. Accuracy of JNET patterns (2B, 2A vs UCAN), pit patterns (Vi and II-IV for UCAN), and macroscopic color should be separately tested and reported, as well as in combination (pp V_i and redness).
The criteria used for EUS diagnosis of submucosal invasion of UCAN need to be specified, because the criterion (“thickening of 3^rd layer”) differs from criteria for sm-invasion described in ref. #6: Kobayashi et al describe iso[93%]-/hypo[7%]-echoic mass lesions causing “narrowing”(=slightly invasive) or “disruption”(= deeply invasive) of hyperechoic 3^rd layer (= submucosa). Despite inconsistent EUS-criteria, the calculation of 90% agreement of EUS depth assessment with histopathology is misleading, when the objective criterion (pathological sm-invasion) was present in only one of ten cases studied. However, the EUS-criterion was positive in 2 of 10 cases, i.e. the chances are 50% for correct or wrong EUS assessment. Overall, the sample size is too small for statistical conclusions.

Minor Points:

The Conclusions differ at end of Discussion (a) and in Abstract (b): Optical diagnosis is ambiguously mentioned in (a), but not at all in (b) inspite of being the primary objective in the Abstract.
pg.1, line 47: …, redness and a VI pit pattern … correct spelling: a V_i pit pattern … (also in Tab.2).
Introduction (pg 2, lines 69-72): “… the indications for endoscopic submucosal dissection (ESD) are limited to lesions no more advanced than low-grade dysplasia (LGD) that meet the following three criteria: .. “. ..
This is not consistent with the malignant UCAN (HGD, T is) listed in Tab. 2, and is not consistent with CQ 21 (and CQ 14) of the JSCCR guideline 2024 (ref. #3). More advanced UCAN (HGD, T is, T1b-superficial) are indications for diagnostic ESD (under preconditions no. 1 - 3), and LGD in ESD specimen no longer is an indication for proctocolectomy. This should be correctly stated.
Please include a short paragraph at end of Results with final outcome (diagnostic ESD and proctocolectomy or segmental resection, curative ESD, resection for AE) of the 23 ESD cases.

Recommendation:

The manuscript requires major revision for acceptance.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Overall assessment and significance

This manuscript presents a retrospective, single-center experience (181 patients, 212 lesions in UC-affected mucosa) evaluating: (1) the diagnostic performance of endoscopic criteria to differentiate ulcerative colitis–associated neoplasia (UCAN) from non-UCAN lesions, (2) the concordance of endoscopic ultrasonography (EUS) with pathological invasion depth, and (3) the concept of endoscopic submucosal dissection (ESD) as a “total biopsy” enabling comprehensive histologic assessment beyond standard forceps biopsies. The clinical question is important: UCAN frequently presents as flat, ill-demarcated lesions on a background of chronic inflammation and fibrosis, where sampling error and diagnostic discordance between biopsy and final pathology are well recognized and have major consequences for management (colectomy/proctocolectomy vs organ-preserving strategies). However, the key conclusions rely on very small numbers of events and evaluable cases (UCAN n=8 overall; ESD n=23; EUS n=10), which limits the stability of diagnostic estimates and makes multivariable modeling particularly fragile (very wide confidence intervals, high risk of overfitting). In its current form, the study is best positioned as a proof-of-concept / expert-center experience demonstrating feasibility and suggesting a potentially useful algorithm, but it requires substantial revisions to methodology reporting, statistical presentation, and tone of inference to avoid overinterpretation. Recommendation: Major revision.

Major strengths

1. Clear clinical gap and rationale: the difficulty of detecting and delineating UCAN in inflamed mucosa, limitations of biopsy, and management implications are well articulated.

2. Coherent institutional strategy: lesions <2 cm treated with EMR/CSP/HSP; lesions ≥2 cm with ESD; additionally, stepwise four-quadrant biopsies around lesions prior to ESD—clinically relevant and practical.

3. Honest presentation of diagnostic asymmetry: high sensitivity/NPV accompanied by low PPV (overcalling UCAN due to regenerative/inflammatory changes), and the Discussion acknowledges this.

4. Useful case-mix transparency: Table 2 provides lesion-level detail (morphology, pit pattern, JNET, Takabayashi classification, and EUS where available), which is valuable for readers.

Major comments

1. Category definitions, mixed “axes,” and risk of verification/work-up bias

The manuscript effectively uses two different classification axes but does not keep them sufficiently separated. Axis A (preoperative clinical/endoscopic hypothesis guiding management): “UCAN-suspected” vs “sporadic-suspected” based on macroscopic appearance, pit pattern, JNET, color, etc. This axis determines escalation to ESD ± EUS vs conventional endoscopic resection. Axis B (reference standard / ground truth): final histopathology “UCAN” vs “non-UCAN” (tubular adenoma, SSL, hyperplastic polyp, inflammatory polyp). A central methodological issue is that the reference standard is not applied uniformly: en bloc ESD specimens (full architecture, margins) are disproportionately obtained in “UCAN-suspected” lesions, whereas “sporadic-suspected” lesions are often verified by conventional resection and/or limited sampling. In UC, this differential verification increases the risk that “no UCAN identified” in the conventionally managed group may partly reflect incomplete verification (sampling error) rather than true absence, thereby inflating performance metrics when reported for the full cohort.

Actionable revisions

Explicitly describe in Methods/Results how selection to ESD and EUS occurred and acknowledge the resulting verification/work-up bias. Reframe diagnostic claims as performance within the center’s algorithm rather than as a universal diagnostic validation. Consider rewording key claims from “endoscopic criteria distinguish UCAN from sporadic neoplasia with high accuracy” to a safer formulation.

2. “100% sensitivity and 100% NPV” are overinterpreted; confidence intervals are mandatory

The reported sensitivity 100% and NPV 100% are based on only eight UCAN cases in the full cohort, and a single misclassification would markedly change estimates. Reporting point estimates without 95% confidence intervals is misleading in this setting.

Actionable revisions:

Add exact 95% confidence intervals (e.g., Clopper–Pearson) for sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy. Given the small number of UCAN events, exact Clopper–Pearson intervals are recommended to avoid overly optimistic estimates and to ensure statistically robust reporting of diagnostic performance. In addition, report UCAN prevalence in the cohort and consider providing likelihood ratios (LR+ = Positive Likelihood Ratio and LR− = Negative Likelihood Ratio) with corresponding confidence intervals. Finally, temper the wording in the Abstract and Conclusions (e.g., “perfect sensitivity in this cohort”) by explicitly acknowledging the wide confidence intervals resulting from the limited number of UCAN cases.

3. Logistic regression in ESD cases: high overfitting risk, unstable ORs, and unclear coding direction

The multivariable logistic regression is performed within the ESD subset (n=23) with ~8 UCAN events, while including many predictors (sex, duration, Mayo Endoscopic Subscore, location, size, morphology, color, pit pattern, Japan NBI Expert Team classification, Takabayashi classification, etc.). This violates standard stability heuristics (very low events-per-variable), producing wide confidence intervals and unstable coefficients. Moreover, the directionality/coding is confusing in places (e.g., color appears as “Reddish = 1.000” reference while the text states “reddened appearance was an independent predictor,” which is not interpretable unless coding/reference categories are fully specified).

Actionable revisions:

Either reduce the model to a parsimonious set of 2–3 pre-specified predictors, or apply penalized regression (e.g., Firth) appropriate for small samples. Clearly report the number of UCAN vs non-UCAN events used in the model and provide unambiguous coding/reference categories. Replace “independent predictors” language with “features associated with UCAN” unless robust modeling is feasible. If a model is retained, provide at least descriptive discrimination metrics (e.g., area under the receiver operating characteristic curve = AUC) while acknowledging limitations.

4. EUS depth staging: internal inconsistency and need for a 2×2 table

The EUS section contains a logical inconsistency: two lesions reportedly had third-layer thickening (suggesting ≥T1b), one of which was T1b (Case 9) and one Tis (Case 10). Yet the text states that EUS was concordant in all but one case (Case 9) with 90% accuracy. If EUS truly suggested ≥T1b in Case 9 and pathology confirmed T1b, Case 9 should be concordant, whereas Case 10 would represent over-staging (false positive). Alternatively, if Case 9 was under-staged by EUS, that needs explicit clarification.

Actionable revisions:

Please provide a clear 2×2 contingency table directly comparing endoscopic ultrasonography (EUS)–based depth assessment with the final histopathological diagnosis, using a clinically meaningful dichotomization of invasion depth into T1a-or-shallower (including Tis) versus ≥T1b (submucosal invasion). The table should explicitly indicate the number of true positives, false positives, false negatives, and true negatives, and the specific cases corresponding to false-positive and false-negative classifications should be clearly identified and described in the text. The reported diagnostic accuracy should then be recalculated to ensure it is numerically consistent with the actual number of discordant cases. In addition, the manuscript should clearly define the EUS imaging criteria used to classify a lesion as ≥T1b, such as disruption of the layered wall structure, thickening or narrowing of the third (submucosal) layer, or irregular submucosal echotexture.

5. Terminology: “sporadic neoplasia/adenoma/lesion” is used interchangeably without definition

The terminology used to describe non-UC-associated lesions requires clarification and greater consistency. Terms such as “sporadic neoplasia,” “sporadic adenoma,” and “sporadic lesion” appear to be used interchangeably without an a priori definition. In UC, the term “sporadic” implies an inflammation-independent carcinogenic pathway that cannot be conclusively established within chronically inflamed mucosa. In this study, “sporadic” seems to function as a heterogeneous, retrospectively defined residual category (non-UCAN: tubular adenomas, SSLs, hyperplastic polyps, inflammatory polyps) rather than a distinct pathogenetic entity.

Actionable revisions:

Define terminology explicitly in Methods. Consider replacing “sporadic” with a more descriptive, conceptually neutral term:
“non-UCAN lesions”.

6. “ESD as total biopsy” is compelling but requires safety/quality metrics and clinical impact data

The “total biopsy” concept is clinically attractive, but the manuscript provides limited procedural outcome data. Standard ESD reporting elements are missing or insufficiently detailed: en bloc resection rate, R0 rate (lateral/deep), curative resection criteria, procedure time, bleeding/perforation details, length of stay, and conversions to surgery. Without these, the argument risks appearing as opinion rather than evidence. Equally important, the manuscript does not quantify decision impact: how often ESD changed management (avoided colectomy, triggered escalation to surgery, clarified “suspected UCAN” as non-UCAN), which is arguably the most clinically meaningful endpoint.

Actionable revisions:

Add standard ESD outcomes (at minimum: en bloc, R0, AEs, procedure time, LOS). Add a brief “clinical impact” summary: proportion avoiding colectomy due to total biopsy, proportion referred to surgery based on ESD pathology, etc.

Follow-up is underreported despite the metachronous risk in UC

UC-as7. sociated neoplasia is not a single-lesion problem; multifocal and metachronous lesions are central to clinical decision-making. The Discussion acknowledges this risk, but the manuscript lacks follow-up data after ESD.

Actionable revisions:

Provide median follow-up duration after ESD (and range), and any available outcomes (metachronous dysplasia/UCAN, colectomy after ESD, progression). Even descriptive reporting would meaningfully strengthen the clinical interpretation.

Minor comments

1. Statistics/Methods wording: the “parametric vs nonparametric” description is generic; specify the tests used, assumptions/ normality checks, and add CIs for diagnostic metrics.

2. Terminology: “US-based depth assessment” in Conclusions may be confused with transabdominal US; please standardize to EUS-based.

3. Figure 1: consider explicitly annotating that no UCAN was identified among lesions <2 cm treated with EMR/CSP/HSP (core finding).

4. Table 2 / inflammation severity: since MES is analyzed in regression, add lesion-level inflammatory activity (segmental MES) in Table 2 or Supplementary material.

5. Peri-lesional biopsies: clarify the protocol (distance from lesion, number of levels/biopsies) and relationship to surveillance standards.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This study addresses a practical challenge in ulcerative colitis–associated neoplasia (UCAN), which often appears flat with ill-defined margins on endoscopy and is difficult to diagnose accurately with conventional biopsy. The authors integrate macroscopic morphology, pit pattern assessment, and the Japan NBI Expert Team (JNET) classification, evaluate the accuracy of endoscopic ultrasonography (EUS) for estimating invasion depth, and further explore the concept of using endoscopic submucosal dissection (ESD) as a “total biopsy” to obtain comprehensive histology. The results suggest that preoperative endoscopic evaluation has strong rule-out performance for UCAN (NPV 100%); EUS shows good concordance for distinguishing T1a from T1b invasion (90%); and, among ESD cases, a reddened appearance and a type VI pit pattern are associated with UCAN. Nevertheless, the retrospective design introduces selection bias and the sample size is limited (particularly for UCAN and the EUS subset). In addition, the claim that “ESD as a total biopsy” is advantageous remains largely conceptual, as there is no direct comparison with conventional multi-site biopsy, targeted biopsy, or piecemeal resection strategies, nor is the clinical decision-making benefit quantified. Strengthening the statistical presentation, definitions, and the clinical decision framework would improve the credibility and generalizability of the conclusions.

Specific comments:

Study population and selection bias: clarification of the inclusion pathway is needed.
It is unclear how the 212 “neoplastic lesions” arising in UC-affected mucosa were identified from the surveillance colonoscopy population. Were all resectable lesions consecutively included, or only those already selected for endoscopic treatment? Beyond lesion size, did the criteria for choosing different modalities (ESD vs EMR/CSP/polypectomy) incorporate factors such as suspected UCAN, indistinct margins, or a flat morphology? Such factors could enrich UCAN cases in the ESD cohort and compromise the robustness of diagnostic performance estimates and risk factor analyses. The authors are encouraged to provide a STROBE-compliant flow diagram clearly outlining the pathway from screening/surveillance to lesion detection, preoperative classification (suspected UCAN or not), EUS implementation, selection of ESD/EMR/CSP, and the final histopathological outcomes.
EUS-based invasion depth assessment: limited sample size and the interpretation criteria require further specification.
The EUS analysis includes only 10 lesions, drawn from the “suspected UCAN” subset, which introduces substantial selection/verification bias; therefore, conclusions that EUS “helps determine optimal treatment strategies” should be phrased more cautiously. Moreover, the criterion that “third-layer thickening/narrowing suggests T1b or deeper invasion” should be defined more explicitly and reported in a reproducible manner, including probe type and frequency (e.g., miniprobe and MHz), examination technique (water immersion vs balloon), operator experience, number of readers, whether interpretation was blinded, and interobserver agreement (e.g., κ statistics if available). Providing more objective descriptors (e.g., thickness thresholds or representative imaging criteria) for “third-layer thickening” would further aid reproducibility. The authors should consider reframing the conclusion as “preliminary/feasible” and emphasize the need for validation in larger cohorts.
Regression analysis in the ESD cohort: inadequate event numbers raise concerns about model stability.
The ESD cohort includes only 23 lesions, with 8 UCAN events. Fitting multivariable models with numerous covariates (sex, disease duration, MES, location, size, morphology, color tone, pit pattern, JNET, Takabayashi classification, etc.) risks severe overfitting due to an insufficient events-per-variable ratio, leading to unstable estimates and potentially chance findings. The authors should clarify the variable selection strategy (univariable screening, clinical priors, etc.) and consider more appropriate approaches for small samples, such as restricting the number of predictors (focusing on 1–2 key variables), using Firth penalized logistic regression, or applying LASSO as an exploratory selection method. Reporting odds ratios with 95% confidence intervals (not only P values) is recommended. The term “independent predictors” should be softened to “exploratory associations” to avoid overinterpretation.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The revised manuscript (available to this reviewer) does not contain any Figs., i.e. an exemplary M-NBI & M-crystal violet-CE for JNET & Kudo-PP & hrEUS of flat-invasive UCAN;

an M-NBI & M-CE for JNET and Kudo-PP & hrEUS of non-invasive UCAN;

an M-NBI & M-CE for JNET and Kudo PP & hrEUS of sporadic NPL (adenoma).

Such Images would much improve the appeal of this analysis to the expert readership.

Thank you for successful major revision of this manuscript.

Author Response

The revised manuscript (available to this reviewer) does not contain any Figs., i.e. an exemplary M-NBI & M-crystal violet-CE for JNET & Kudo-PP & hrEUS of flat-invasive UCAN; an M-NBI & M-CE for JNET and Kudo-PP & hrEUS of non-invasive UCAN; an M-NBI & M-CE for JNET and Kudo PP & hrEUS of sporadic NPL (adenoma). Such Images would much improve the appeal of this analysis to the expert readership.　Thank you for successful major revision of this manuscript.

Response: We once again thank you for the time and effort dedicated to reviewing our manuscript. We apologize for the oversight; representative images for each category have now been provided as Figures 1, 2, and 3.

Specifically, we have added representative examples of flat-invasive UCAN (Figure 3), non-invasive UCAN (Figure 4), and sporadic non-polypoid lesions (adenoma) (Figure 5) to the Results section.

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript addresses an important and clinically relevant challenge in the management of ulcerative colitis-associated neoplasia (UCAN), namely the difficulty of accurate preoperative diagnosis in an inflammation-altered mucosal environment. The study is well designed, includes a clearly defined cohort (212 lesions in 181 patients), and provides a structured evaluation of endoscopic features, EUS-based depth assessment, and the role of ESD.
The revised version has substantially improved in methodological transparency and statistical reporting. In particular, the inclusion of confidence intervals, a 2×2 contingency table for EUS performance, and explicit reporting of diagnostic metrics significantly strengthens the analytical rigor. The authors appropriately acknowledge the limited sample size and clearly position the EUS findings as exploratory, which improves the credibility of the conclusions.
A key strength of the study is the conceptual framing of ESD as a “total biopsy,” providing comprehensive histological assessment and supporting clinical decision-making in a setting where conventional biopsy is often insufficient. This perspective adds meaningful clinical value and aligns well with current trends toward organ-preserving strategies in inflammatory bowel disease.
The Discussion is balanced and well contextualized within the existing literature, and the limitations—including potential verification bias and single-center design—are appropriately addressed. Minor revisions are suggested to further improve clarity and tone: slightly soften the wording of the conclusions to reflect the exploratory nature of the findings, explicitly emphasize the impact of diagnostic ESD on clinical decision-making (e.g., avoidance of unnecessary proctocolectomy). Consider adding a brief statement highlighting the need for multicenter validation.
Overall, this is a clinically meaningful and methodologically sound study that will be of interest to gastroenterologists and endoscopists managing IBD-related neoplasia.

Author Response

Response to Reviewer 2

Response: We once again thank you for the time and effort dedicated to reviewing our manuscript and for your positive consideration of our manuscript. We also appreciate your insightful comments. We have accordingly added the following to the revised manuscript:

“An integrated approach that combines optical diagnosis, EUS-based depth assessment, and diagnostic ESD may enhance the accurate differentiation of UCAN from sporadic neoplasia. In particular, diagnostic ESD can diagnostic ESD offers comprehensive histological evaluation of the entire lesion and may support clinical decision-making, including the potential avoidance of unnecessary total proctocolectomy in selected cases.

However, given the retrospective design and the limited number of cases, especially in the EUS analysis, these findings should be interpreted as exploratory. Further validation through larger, prospective multicenter studies is warranted to confirm the clinical utility of this stepwise strategy.”