Aggressive colorectal carcinoma phenotypes of invasion can be assessed reproducibly and effectively predict poor survival: interobserver study and multivariate survival analysis of a prospectively collected series of 299 patients after potentially curative resections with long-term follow-up.

AIMS

To test whether assessment of colorectal carcinoma phenotypes of invasion can be taught successfully to pathologists not familiar with these features and to confirm their prognostic impact.

METHODS AND RESULTS

A junior member of staff was supplied with written information and a generous set of images on how to type the invasive margin of colorectal carcinomas (expansive versus infiltrative) and how to assess tumour budding (counting on pan-cytokeratin immunostains; cut-off at the 66th percentile). An interobserver study yielded kappa values of 0.578 and 0.438 for typing the invasive margin and assessment of tumour budding, respectively. Margin typing improved significantly to κ=0.939 after a training session. However, using a cut-off, assessment of tumour budding improved only moderately to κ=0.629 although, numerically, divergences were within ±10%. On univariate analysis, a high-degree of tumour budding and the infiltrative type of invasive margin were strong negative prognostic factors. The Cox model included nodal status, tumour budding, serosal penetration and venous angioinvasion. Importantly, using the tumour budding counts as a numerical variable for the Cox model also yielded significant odds, allowing the constraints of a cut-off to be relinquished.

CONCLUSIONS

Assessment of colorectal carcinoma phenotypes of invasion can be learnt and performed with confidence, and their prognostic impact is confirmed in this independent series.