Diagnostic concordance of reporting lymphovascular invasion in breast cancer

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

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.

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.

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.