Departments of *Pathology¶Biostatistics, Vanderbilt University, Nashville, TN†Department of Pathology, MUSC Medical Center, Charleston, SC‡Department of Pathology, Cedars-Sinai Medical Center, Los Angeles, CA§Department of Pathology, Massachusetts General Hospital, Boston, MA∥Department of Pathology, Dartmouth Hitchcock Medical Center, Lebanon, NH
Checking for direct PDF access through Ovid
BackgroundLymphovascular invasion (LVI) in colorectal cancer (CRC) is considered a strong stage-independent prognostic factor and influences decisions regarding adjuvant chemotherapy in patients with stage II tumors. However, the degree of interobserver agreement among pathologists for LVI in CRC is largely unknown. This study was undertaken to examine such interobserver variability, and we hypothesized that the use of immunohistochemical markers for vascular and lymphatic channels could improve interobserver agreement.DesignFifty cases of American Joint Committee on Cancer stage II moderately differentiated CRC from 1990 to 2005 from the pathology archives were selected; mucinous, medullary, and other recognized special subtypes were excluded. Fifty hematoxylin and eosin (H&E) slides (1 from each case) were circulated to 6 gastrointestinal pathologists, who independently assessed small and large vessel invasion. No diagnostic guidelines were given to the participating pathologists; each was instructed to apply the criteria for LVI that he or she used in daily practice. Immunohistochemistry (IHC) for D2-40 and CD31 was performed on corresponding paraffin blocks. The IHC slides were randomized, recirculated, and rescored for LVI. Results were analyzed by kappa (κ) statistics, which correct for agreement by chance, and for percentage agreement.ResultsThe average κ values were determined for the H&E slides (large and small vessel), CD31 (small vessel), and D2-40 (small vessel) (Fig. 1). Agreement was fair for H&E small vessel invasion [κ=0.28; 95% confidence interval (CI): 0.22-0.34]. The least agreement was seen in interpretation of H&E large vessel invasion (κ=0.18; 95%CI: 0.11-0.26). Agreement was not improved by use of immunohistochemical stains: CD31 (large vessel, κ=0.42, 95%CI: 0.20-0.63, small vessel, κ=0.26, 95%CI: 0.10-0.42) and D2-40 (κ=0.32, 95%CI: 0.21-0.42).ConclusionsInterobserver variability in diagnosis of LVI was substantial on H&E slides and did not improve upon use of IHC. Agreement in evaluation of large vessel invasion was only slightly higher than would be seen by chance alone. This study highlights the need for criteria in evaluation of LVI, as this assessment may impact patient prognosis and thus change the course of clinical treatment.