From the AGA Journals

Lesion morphology drives optical evaluation’s accuracy for predicting SMIC

View on the News

Better prediction for some lesions

Because endoscopists are becoming more proficient with endoscopic mucosal resection (EMR) and are pushing the bounds with endoscopic submucosal dissection (ESD), there is a need for high-quality endoscopic markers of submucosal invasion (SMI) to help guide decision-making for management.

Rehman Sheikh, MD, is a gastroenterologist at the Baylor College of Medicine in Houston.

Dr. Rehman Sheikh

This study by Dr. Vosko and colleagues demonstrated a high degree of accuracy in predicting SMI via optical evaluation for a select group of large nonpedunculated colorectal polyps (LNPCP). The features associated with SMI were Kudo Pit Pattern V, ulceration, depression (Paris 0-IIc morphology), and rigidity or fixation.

The authors demonstrated that optical evaluation was highly accurate in detecting submucosal invasion for flat LNPCP with a high sensitivity and specificity. The sensitivity was considerably lower when evaluating nodular LNPCPs with a higher miss rate in polyps >4.0 cm and those located in the rectosigmoid colon. Of the endoscopic features assessed, Kudo pit pattern had the highest reliability in predicting SMI.

These data further tip the scale in favor of EMR as the appropriate therapeutic option for flat LNPCP in absence of features of SMI outlined by the authors. It also highlights the need for all endoscopists to be well versed in Kudo Pit classification and proficient in assessing for rigidity, fixation, and depression as the therapeutic decision (namely EMR vs. ESD vs. surgery) is often made by the endoscopist discovering the polyp. More studies are needed to identify endoscopic characteristics that provide a high sensitivity and specificity for SMI in nodular LNPCPs.

Rehman Sheikh, MD, is a gastroenterologist at the Baylor College of Medicine in Houston. He has no conflicts to declare.


The diagnostic performance of optical evaluation for submucosal invasive cancer (SMIC) in patients with large (≥20 mm) nonpedunculated colorectal polyps (LNPCPs) may be dependent on lesion morphology. While optical evaluation featured excellent performance in the assessment of flat lesions, the assessment only featured decent performance in nodular lesions, underscoring the need for additional evaluation algorithms for these lesions.

Endoscopists rely on the accuracy of real-time optical evaluation to facilitate appropriate selection of treatment; however, in studies focusing on LNPCPs, the performance of optical evaluation is modest.

The stratification of optical evaluation by lesion morphology may enable more accurate “implementation of a selective resection algorithm by identifying lesion subgroups with accurate optical evaluation performance characteristics,” first authors Sergei Vosko, MD, and Neal Shahidi, MD, of the department of gastroenterology and hepatology, Westmead Hospital, Sydney, and colleagues wrote in Clinical Gastroenterology and Hepatology.

Given the potential importance of stratification in optical evaluation, Dr. Vosko and colleagues assessed the performance of the optical assessment modality based on lesion morphology in a prospective cohort of 1,583 LNPCPs measuring at least 20 mm in patients (median age, 69 years) referred for endoscopic resection.

In the observational cohort, centers performed optical evaluation before endoscopic resection. The optical prediction of SMIC was based on several different established features, including Kudo V pit pattern, depressed morphology, rigidity/fixation, and ulceration. The researchers calculated optical evaluation performance outcomes, which were reported by the dominant morphology, namely nodular (Paris 0–Is/0– IIaDIs) versus flat (Paris 0–IIa/0–IIb).

Across the overall cohort, the median lesion size was 35 mm. The investigators identified a total of 855 flat LNPCPs and 728 nodular LNPCPs, with 63.9% of LNPCPs considered granular. Additionally, the researchers reported submucosal invasive cancer in 146 LNPCPs (9.2%).

According to the investigators, the overall sensitivity of optical evaluation to diagnose submucosal invasive cancer was 67.1% (95% confidence interval, 59.2%-74.2%), while the overall specificity was 95.1% (95% CI, 93.9%-96.1%). The investigators reported significant differences between flat vs. nodular LNPCPs in terms of sensitivity (90.9% vs. 52.7%, respectively; P <.001) and specificity (96.3% vs. 93.7%; P =.027).

Overall, the SMIC miss rate was 3.0% (95% CI, 2.3%-4.0%). There was a significant difference in the SMIC miss rate between flat and nodular LNPCPs (0.6% vs. 5.9%, respectively; P < .001).

Independent predictors of missed SMIC on optical evaluation, as identified in the multiple logistic regression analysis, included nodular morphology (odds ratio, 7.2; 95% CI, 2.8-18.9; P < .001), rectosigmoid location (OR, 2.0; 95% CI, 1.1-3.7; P =.026), and size of at least 40 mm (OR, 2.0; 95% CI, 1.0-3.8; P =.039).

Based on the findings, the researchers suggested that all flat lesions, in the absence of optical features consistent with submucosal invasive cancer, should subsequently be removed by high-quality endoscopic mucosal resection, in conjunction with the application of “site-specific modifications and ancillary techniques where needed.”

One limitation of this study is how lesion morphology was classified, which can in some cases be subjective.

The researchers added that additional refinement is required “to robustly apply a selective resection algorithm irrespective of lesion morphology” given the modest performance value of optical evaluation in nodular lesions. “Nevertheless, it is imperative that all endoscopists embrace optical evaluation in everyday clinical practice, thus harnessing its proven ability to influence resection technique selection and the associated clinical and economic ramifications,” they concluded.

The study received financial support the Cancer Institute of New South Wales, in addition to funding from the Gallipoli Medical Research Foundation. One author reported receiving research support from Olympus Medical, Cook Medical, and Boston Scientific. The remaining authors disclosed no conflicts.

Next Article: