Large sessile or flat colorectal polyps or laterally spreading lesions were most likely to contain covert malignancies when their location was rectosigmoid, their Paris classification was 0-Is or 0-IIa+Is, and they were nongranular, according to the results of a multicenter prospective cohort study of 2,106 consecutive patients reported in the September issue of Gastroenterology ( doi: 10.1053/j.gastro.2017.05.047 ).

“Distal nongranular lesions have a high risk of occult SMIC [submucosal invasive cancer], whereas proximal, granular 0-IIa lesions, after a careful assessment for features associated with SMIC, have a very low risk,” wrote Nicholas G. Burgess, MD, of Westmead Hospital, Sydney, with his associates. “These findings can be used to inform decisions [about] which patients should undergo endoscopic submucosal dissection, endoscopic mucosal resection, or surgery.”

Many studies of colonic lesions have examined predictors of SMIC. Nonetheless, clinicians need more information on factors that improve clinical decision making, especially as colonic endoscopic submucosal dissection becomes more accessible, the researchers said. Large colonic lesions can contain submucosal invasive SMICs that are not visible on endoscopy, and characterizing predictors of this occurrence could help patients and clinicians decide between endoscopic submucosal dissection and endoscopic mucosal resection. To do so, the researchers analyzed histologic specimens from 2,277 colonic lesions above 20 mm (average size, 37 mm) that lacked overt endoscopic high-risk features. The study ran from 2008 through 2016, study participants averaged 68 years of age, and 53% were male. A total of 171 lesions (8%) had evidence of SMIC on pathologic review, and 138 lesions had covert SMIC. Predictors of overt and occult SMIC included Kudo pit pattern V, a depressed component (0-IIc), rectosigmoid location, 0-Is or 0-IIa+Is Paris classification, nongranular surface morphology, and larger size. After excluding lesions with obvious SMIC features – including serrated lesions and those with depressed components (Kudo pit pattern of V and Paris 0-IIc) – the strongest predictors of occult SMIC included Paris classification, surface morphology, size, and location.

“Proximal 0-IIa G or 0-Is granular lesions had the lowest risk of SMIC (0.7% and 2.3%), whereas distal 0-Is nongranular lesions had the highest risk (21.4%),” the investigators added. Lesion location, size, and combined Paris classification and surface topography showed the best fit in a multivariable model. Notably, rectosigmoid lesions had nearly twice the odds of containing covert SMIC, compared with proximal lesions (odds ratio, 1.9; 95% confidence interval, 1.2-3.0; P = .01). Other significant predictors of covert SMIC in the multivariable model included combined Paris classification, surface morphology (OR, 4.0; 95% CI, 1.2-12.7; P = .02), and increasing size (OR, 1.2 per 10-mm increase; 95% CI, 1.04-1.3; P = .01). Increased size showed an even greater effect in lesions exceeding 50 mm.

Clinicians can use these factors to help evaluate risk of invasive cancer in lesions without overt SMIC, the researchers said. “One lesion type that differs from the pattern is 0-IIa nongranular lesions,” they noted. “Once lesions with overt evidence of SMIC are excluded, these lesions have a low risk (4.2%) of harboring underlying cancer.” Although 42% of lesions with covert SMIC were SM1 (potentially curable by endoscopic resection), no predictor of covert SMIC also predicted SMI status.

Funders included Cancer Institute of New South Wales and Gallipoli Medical Research Foundation. The investigators had no conflicts of interest.