Comparison of the diagnostic accuracy of lymphatic endothelium markers: Bayesian approach.

Tumor lymphatic density is evaluated by means of specific lymphatic endothelium markers, and is a potential predictor of clinically meaningful outcomes. There are many claims on the postulated superiority of some of these markers to identify lymphatics, always in the absence of quantitative data. We therefore compared the diagnostic accuracy of the antibody against podoplanin and the commercially available D2-40, employing Bayesian statistics to account for the absence of a gold standard. We used the pan-endothelial marker CD34 to identify 23,542 distinct blood and lymphatic vessels in sections from 30 formalin-fixed, paraffin-embedded archival tissue blocks of head and neck squamous cell carcinoma specimens. We stained two adjacent sections with podoplanin and D2-40 and identified the continuum of each stained vessel in the sections with a comprehensive method. Overall, 1,864 vessels were stained with both markers, 119 only with podoplanin and 391 only with D2-40. Significantly more vessels with intraluminal red blood cells were stained with D2-40 compared to podoplanin (McNemar's P<0.0001). Both antibodies had extremely high specificity (99.7% (95% credible interval (CrI): 99.5-99.9%) and 98.8% (95% CrI: 98.3-99.5%) for podoplanin and D2-40, respectively) and very high sensitivity (92.6% (95% CrI: 86.1-97.9%) and 97.3% (95% CrI: 94.9-99.2%) for podoplanin and D2-40, respectively). Inferences were qualitatively similar when we took into account in the analyses the possibility that the two tests (antibodies) may be correlated. We calculated that 96.3% (95% CrI: 94.2-98.6%) of the vessels stained with podoplanin and 88.9% (95% CrI: 83.9-95.7%) of the vessels stained with D2-40 were truly lymphatics. These numbers were in agreement with the observed number of stained vessels without intraluminal red blood cells. Our results suggest that both antibodies are excellent lymphatic endothelium markers and that there may be little reason to prefer either of them in most settings. Modern Pathology (2005) 18, 1490-1497. doi:10.1038/modpathol.3800457; published online 1 July 2005.