Pathologic predictive factors for lymph node metastasis in submucosal invasive (T1) colorectal cancer: a systematic...

Colorectal adenocarcinoma with depth of invasion ≤1,000 μm from the muscularis mucosa and favorable histology is now considered for local resection. We aimed to examine the strength of evidence for this emerging practice.

We searched Medline, Scopus, and Cochrane (1950-2011), then performed a meta-analysis on the risk of lymph node metastasis in nonpedunculated (sessile and nonpolypoid) T1 colorectal cancers. We included studies with nonpedunculated lesions, actual invasion depth, and pathologic factors of interest. Synchronous, polyposis or secondary cancers, and chemoradiation studies were excluded. Our primary outcome was the risk of LNM. We analyzed using Review Manager; we estimated heterogeneity using Cochran Q χ(2) test and I (2). We generated summary risk ratios using a random effects model, performed sensitivity analyses, and evaluated the quality of evidence using GRADEPro.

We identified 209 articles; 5 studies (n = 1213 patients) met the inclusion criteria. The risk of LNM in nonpedunculated ≤1,000 μm is 1.9 % (95 % confidence interval 0.5-4.8 %). The risk for all T1 is 13 % (95 % confidence interval 11.5-15.4 %). Characteristics protective against LNM were ≤1,000 μm invasion, well differentiation, absence of lymphatic and vascular invasion, and absence of tumor budding. We did not detect significant study heterogeneity. The quality of evidence was poor.

Well-differentiated nonpedunculated T1 colorectal cancer invasive into the submucosa ≤1,000 μm, without lymphovascular involvement or tumor budding, has the lowest risk of nodal metastasis. Importantly, the risk was not zero (1.9 %), and the qualitative formal analysis of data was not strong. As such, endoscopic resection alone may be adequate in select patients with submucosal invasive colorectal cancers, but more studies are needed. Overall, the quality of evidence was poor; data were from small retrospective studies from limited geographic regions.