Diagnostic concordance of reporting lymphovascular invasion in breast cancer

Abstract

Aims: This study aims to assess the diagnostic agreement of lymphovascular invasion (LVI) in invasive breast cancer (BC).

Methods: Data on LVI were collected from the UK National Health Service Breast Screening Programme pathology external quality assurance scheme database. 101 BCs assessed over a 10-year period (2004-2014) were included. Cases were scored by an average of 600 pathologists. Three H&E stained slides from each case were reviewed by three pathologists and additional variables were evaluated.

Results: In the whole series, the overall κ value was 0.4 (range 0.26-0.53). On review, LVI was detected in all three slides in 20 cases (20%), in two slides in 12 cases and in one of the three slides in 9 cases and was not seen in 60 cases. For concordance analysis, the first and last groups were used to represent cases with definite (LVI+) and absent LVI (LVI-), respectively. In the LVI+group (n=20), the level of agreement ranged from 0.54 to 0.99 (median 0.86). In the LVI- group (n=60), the level of agreement ranged from 0.52 to 1.00 (median 0.93), with 44% of cases showing interobserver concordance of >95%. There was a correlation between increasing number of involved lymphovascular spaces in the section and higher LVI reporting concordance. Some degree of retraction/fixation artefacts was observed in 35% of cases; this was associated with a lower concordance rate.

Conclusions: The concordance of reporting LVI is variable. Cases without LVI and those with multiple involved vessels are likely to have the highest concordance and the highest detection rates.

Keywords: breast cancer; interobserver agreement; lymphovascular invasion.