Variability in the identification of lymphovascular space invasion for early stage cervical cancer

Objective: To evaluate the inter- and intra-rater variability of lymphovascular space invasion (LVSI) in early stage cervical cancer. Methods: We identified invasive cervical cancer tissue samples from radical hysterectomies in our institutional pathology database. The cases were stained with Hematoxylin & Eosin (H&E) and immunostains (CD-31 and D2-40). They were evaluated for the presence of LVSI by 6 pathologists on 3 separate occasions: with H&E staining only, then with H&E and immunostained specimens, and finally using a shared written criterion for diagnosis of LVSI. With 80 cases, a two-sided 95% confidence interval for the Kappa of 0.7 with a precision of 0.1 on each side was estimated. Results: Stage distribution was: IA 10%, IB 85%, and IIA 5%. The majority of cases were squamous cell carcinoma (55%), followed by adenocarcinoma (39%) and adenosquamous or other histology (6%). The mean inter-rater Kappa was 0.41 (95% CI: 0.37–0.45) for H&E. Usage of immunohistochemistry made a statistically significant improvement in the mean Kappa, but it still remained low: 0.52 (p = 0.02). Adding evaluation criteria for LVSI did not significantly increase the mean Kappa: 0.49 (p = 0.16). The mean intra-rater variability of H&E staining alone compared with H&E staining plus immunostaining was 0.53 (range: 0.43–0.64). The mean Kappa comparing H&E staining and H&E staining with criteria was 0.50 (range: 0.40–0.59). Conclusions: We noted high inter- and intra-rater variability in the diagnosis of LVSI underscoring the challenges of LVSI diagnosis. Considering the significance assigned to LVSI and its implication for treatment, comprehensive guidelines with regards to determination of LVSI status are of paramount importance.