Reproducibility of lymphovascular space invasion (LVSI) assessment in endometrial cancer

Aims Lymphovascular space invasion (LVSI) in endometrial cancer (EC) is an important prognostic variable impacting on a patient's individual recurrence risk and adjuvant treatment recommendations. Recent work has shown that grading the extent of LVSI further improves its prognostic strength in patients with stage I endometrioid EC. Despite this, there is little information on the reproducibility of LVSI assessment in EC. Therefore, we designed a study to evaluate interobserver agreement in discriminating true LVSI from LVSI mimics (Phase I) and reproducibility of grading extent of LVSI (Phase II). Methods and results Scanned haematoxylin and eosin (H&E) slides of endometrioid EC (EEC) with a predefined possible LVSI focus were hosted on a website and assessed by a panel of six European gynaecological pathologists. In Phase I, 48 H&E slides were included for LVSI assessment and in Phase II, 42 H&E slides for LVSI grading. Each observer was instructed to apply the criteria for LVSI used in daily practice. The degree of agreement was measured using the two-way absolute agreement average-measures intraclass correlation coefficient (ICC). Reproducibility of LVSI assessment (ICC = 0.64, P < 0.001) and LVSI grading (ICC = 0.62, P < 0.001) in EEC was substantial among the observers. Conclusions Given the good reproducibility of LVSI, this study further supports the important role of LVSI in decision algorithms for adjuvant treatment.