Premium
International validation of a risk prediction algorithm for patients with malignant colorectal polyps
Author(s) -
Richards C.,
Levic K.,
Fischer J.,
Eglinton T.,
Ramsay G.,
Kumarasinghe P.,
Raftopoulos S.,
Brown I.
Publication year - 2020
Publication title -
colorectal disease
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.029
H-Index - 89
eISSN - 1463-1318
pISSN - 1462-8910
DOI - 10.1111/codi.15365
Subject(s) - medicine , lymphovascular invasion , algorithm , colorectal cancer , receiver operating characteristic , residual , resection margin , disease , margin (machine learning) , population , surgery , radiology , gastroenterology , resection , cancer , machine learning , metastasis , environmental health , computer science
Aim The optimal management strategy for patients with endoscopically resected malignant colorectal polyps (MCP) has yet to be defined. The aim of this study was to validate a published decision‐making tool, termed the Scottish Polyp Cancer Study (SPOCS) algorithm, on a large international population. Methods The SPOCS algorithm allocates patients to risk groups based on just two variables: the polyp resection margin and the presence of lymphovascular invasion (LVI). The risk groups are termed low (clear margin, LVI absent), medium (clear margin, LVI present) or high (involved/non‐assessable margin). The International Polyp Cancer Collaborative was formed to validate the algorithm on data from Australia, Denmark, UK and New Zealand. Results In total, 1423 patients were included in the final dataset. 680/1423 (47.8%) underwent surgical resection and 108/680 (15.9%) had residual disease (luminal disease 8.8%, lymph node metastases 8.8%). The SPOCS algorithm classified 602 patients as low risk (in which 1.5% had residual disease), 198 patients as medium risk (in which 7.1% had residual disease) and 484 as high risk (in which 14.5% had residual disease) ( P < 0.001, χ 2 test). Receiver operating characteristic curve analysis demonstrated good accuracy of the algorithm in predicting residual disease (area under the curve 0.732, 95% CI 0.687–0.778, P < 0.001). When patients were designated as low risk, the negative predictive value was 98.5%. Conclusion The SPOCS algorithm can be used to predict the risk of residual disease in patients with endoscopically resected MCPs. Surgery can be safely avoided in patients who have a clear margin of excision and no evidence of LVI.