Detection of gastric atrophy by circulating pepsinogens: A comparison of three assays

Background: Circulating levels of pepsinogens have been used in high gastric cancer-risk Asian and European populations to triage endoscopic evaluation for more severe pathology. There are different analytic methods with uncertain correlations. We therefore compared diagnostic performance of three commonly used pepsinogen assays to detect histologically confirmed gastric atrophy. Methods: We tested plasma samples from adult patients with (n=50) and without (n=755) moderate or severe gastric corpus atrophy, as determined histologically by consensus of three expert pathologists. A single laboratory measured pepsinogens I (PgI) and II (PgII) using commercially available assays: two ELISA assays produced by Biohit (Finland) and Vector Best (Russia), and a latex agglutination assay from Eiken (Japan). Quantitative correlations were assessed by Spearman statistics. Receiver operating characteristic (ROC) curves vs histological diagnosis were calculated using both the manufacturers' and optimized cutoffs. Results: Pepsinogen levels were highly correlated among the assays (pairwise Rhos: PgI≥0.84, PgII≥0.87; all P-values<.01). Based on manufacturers' cutoffs, sensitivities, specificities and areas under the ROC curve for detecting moderate to severe histological corpus atrophy by PgI/PgII were 44%/91%/0.70, 56%/84%/0.76, and 52%/90%/0.77 for Biohit, Vector Best and Eiken, respectively. Cutoffs optimized by ROC or data mining analyses did not substantially improve test performance. Conclusions: Commercial assays for pepsinogen have good relative agreement but are imperfect tests for clinical diagnosis of gastric atrophy. Impact: Pepsinogen testing alone does not provide sufficient information for gastric cancer risk stratification. Future investigations should focus on other potential markers, in combination with pepsinogens.