Many studies have shown poor reproducibility among pathologists for diagnosing dysplasia in Barrett's esophagus (BE). Immunohistochemical stains (IHC) are not widely used due to overlapping expression patterns in reactive and dysplastic processes. We hypothesized that markers involved in cell-cycle (cyclin D1, Ki-67, P16), differentiation/cell-cell interaction (β-catenin, SATB2 CD44, OCT4) and senescence (γH2AX) would produce different results in reactive and dysplastic processes.
A micrograph album of 40 H&E and matching IHCs depicting optimally oriented lesions were evaluated independently by 3 pathologists. Expression was scored separately in the surface, isthmus, and base regions of the glands.
Statistical analysis showed that surface Ki-67 expression showed the largest difference in expression and smallest P value (P < .001) for identifying dysplasia. At a cutoff level of 5% or less, negative predictive value (NPV) was 100%. κ correlation between pathologists improved from substantial to almost perfect (0.70-0.95) using ancillary surface Ki-67.
A case-control study with glass slides including all diagnostic categories using this parameter confirmed improved κ correlation among pathologists (0.29 vs 0.60), better correlation with outcomes (76% vs 69%), increased odd risks (15.3) for progression in positive cases, and an improvement in sensitivity (88% vs 64%) and NPV (88% vs 73%) compared to histology alone.