Background: Various methods to evaluate immunohistochemical staining (IHC) for the diagnosis of type 1 autoimmune pancreatitis (AIP) have been proposed. Our goal was to determine the most useful IHC method for the diagnosis of AIP.
Methods: Specimens of AIP (18 patients), chronic pancreatitis (CP, 24 patients), and pancreatic ductal adenocarcinoma (PDA, 45 patients) were evaluated with IHC for immunoglobulin G (IgG), IgG1, IgG4, and CD138 (syndecan-1). The number of IHC-positive cells was counted in 3, 5, and 10 different high-power fields (HPFs) by selecting fields with the most numerous positive cells (hotspot) or by randomly selecting fields in the affected areas (random). We evaluated the mean number of IgG4-positive plasma cells (IgG4+)/HPF (mean IgG4+), the number of fields with >10 and >50 IgG4+ (NOF >10 and NOF >50 IgG4+), the ratio of IgG4+/IgG+, IgG4+/IgG1+, and IgG4+/CD138+.
Results: Analysis with receiver operator characteristic curves revealed that accurate and practical parameters in 3 HPFs were mean IgG4+ with the hotspot method (sensitivity, 88.9; specificity, 92.8 %), mean IgG4+ with the random method (100, 95.7 %), and NOF >10 IgG4+ with the random method (94.4, 97.1 %). These results were as accurate as results from 5 HPFs to 10 HPFs. The combination of mean IgG4+ and IgG4+/IgG+ did not provide more accurate diagnosis for AIP than a single criterion itself.
Conclusions: Mean IgG4+ or NOF >10 IgG4+ with the random method in 3 HPFs was a useful and simple diagnostic method for AIP. The combined criteria of mean IgG4+ and IgG4+/IgG+ might not be required for accurate diagnosis of AIP.