Feasibility of new prognostic classification for rectal cancer.

Sixty slides from 60 blocks taken from 30 colonic carcinomas were circulated twice to six histopathologists of varying experience. Five of the six pathologists showed a good to excellent intraobserver agreement for assessment of the character of the invasive margin (0.44 less than kappa less than 0.82), which was not significantly affected by sampling (0.40 less than kappa less than 0.56, comparing both slides from each tumour) or observer (five of six pathologists agreeing on 46 of 60 slides). Pathologists were unreliable in assessing peritumoural lymphocytic infiltrates, with only two pathologists achieving moderate levels of intraobserver agreement (-0.03 less than kappa less than 0.52). The interobserver agreement for peritumoural lymphocytic infiltrates was also low (kappa less than 0.29) between the three most experienced pathologists. The assessment of peritumoural lymphocytic infiltrates was significantly affected by sampling, the two pathologists with the lowest intraobserver variation achieving kappa values of 0.21 and 0.10 between the 30 paired slides from each tumour. The character of the invasive margin was reliably assessed, was not dependent on sample, and added useful prognostic information, but peritumoural lymphocytic infiltration is not a reproducible observation and may therefore not add useful prognostic information in routine use.